Appendix J. Modeling Emission Inventory

Transcription

1 Appendix J Modeling Emission Inventory

2 Modeling Emission Inventory for the PM2.5 State Implementation Plan in the San Joaquin Valley Prepared by California Air Resources Board San Joaquin Valley Air Pollution Control District Prepared for United States Environmental Protection Agency Region IX June 19,

3 Contents 1.1. Inventory Coordination Background Inventory Years Base Case Modeling Inventory (2013) Reference Year Modeling Inventory (2013) Future Year Modeling Inventory (2020/2024/2025) Spatial Extent of Emission Inventories Estimation of Base Year Modeling Inventory Terminology Temporal Distribution of Emissions Monthly Variation Weekly Variation Daily Variation Spatial Allocation Spatial Allocation of Area Sources Spatial Allocation of Point Sources Spatial Allocation of Wildfires, Prescribed Burns and Wildland Fire Use Spatial Allocation of Ocean going vessels (OGV) Spatial Allocation of On-road Motor Vehicles Spatial Allocation of Biogenic Emissions

4 2.3.7 Spatial Allocation of Residential Wood Combustion Emissions Speciation Profiles Methodology for Developing Base Case, Baseline, and Future Projected Emissions Inventories Surface Temperature and Relative Humidity Fields Insolation Effects Estimation of Gridded Area and Point sources Estimation of On-road Motor Vehicle Emissions General Methodology ITN Activity Data Spatial Adjustment Temporal Adjustment (Day-of-Week adjustments to EMFAC daily totals) Temporal Adjustment (Hour-of-Day re-distribution of hourly travel network volumes) Summary of On-road Emissions Processing Steps Adjustment to the Future Year On-road Emissions Estimation of Gridded Biogenic Emissions Estimation of Ocean-going Vessel (OGV) Emissions Estimation of Other Day-Specific Sources Wildfires and Prescribed Burns Paved Road Dust: Unpaved Road Dust:

5 3.7.4 Agricultural Burning: Residential Wood Curtailment Estimation of Agricultural Ammonia Emissions: Closed Facilities Application of Control Measure Reduction Factors in San Joaquin Valley Charbroiling Residential Wood Combustion Future Year Future Years 2024 and Quality Assurance of Modeling Inventories Area and Point Sources Area and Point Sources Temporal Profiles On-road Emissions Day-specific Sources Wildfires and Prescribed Burns Paved Road Dust Unpaved Road Dust Agricultural Burning Additional QA Model ready files QA Bibliography

6 Appendix A: Day of week redistribution factors by vehicle type and county Appendix B: Hour of Day Profiles by vehicle type and county Appendix C: Scaling procedures after DTIM processing Appendix D: Additional temporal profiles

7 Table of Figures Figure 1 Spatial coverage and parameter summary of modeling domains Figure 2 Wilderness mask on ca-4km state domain Figure 3 Example of population surrogate before (a) and after (b) wilderness mask application Figure 4 New spatial surrogate of fireplaces based on housing population (2017) Figure 5 Map of residential wood curtailment areas Figure 6 Block diagram for on-road processing Figure 7 Hot spot areas in San Joaquin Air Basin Figure 8 Hot Spot Areas for application of under-fired charbroiling PM2.5 reductions.. 64 Figure 9 Example of a spatial plot by source category Figure 10 Screen capture of a SMOKE-generated QA report Figure 11 Screenshot of comparison of inventories report Figure 12 Daily variation of NOx emissions for mobile sources for San Luis Obispo

8 List of Tables Table 1 Modeling domain parameters Table 2 Inventory terms for emission source types Table 3 Day of week variation factors Table 4 Daily variation factors Table 5 Spatial Surrogates Table 6 List of Surrogates in which the wilderness mask was applied Table 7 Vintage of travel demand models for link based and traffic analysis zone Table 8 DTIM Emission Categories Table 9 Vehicle classification and type of adjustment Table 10 County specific factors for Table 11 County specific factors for Table 12 County specific burn cleaner retention factors Table 13 NOx and PM2.5 other sources reduced Table 14 Day of week adjustment by vehicle class and county Table 15 Hour of Day Profiles by vehicle type and county Table 16 Day of week temporal profiles from the Agricultural Emissions Temporal and Spatial Allocation Tool (AgTool) Table 17 Daily temporal profiles from the Agricultural Emissions Temporal and Spatial Allocation Tool (AgTool)

9 Development of PM2.5 Emissions Inventories Emission inputs for air quality modeling (commonly and interchangeably referred to as modeling inventories or gridded inventories ) have been developed by CARB and district staff. These inventories support multiple SIPs across California to address nonattainment of the federal PM2.5 (particulate matter 2.5µ in diameter and smaller) standards. CARB maintains an electronic database of emissions and other useful information to generate aggregate emission estimates at the county, air basin and district level. This database is called the California Emission Inventory Development and Reporting System (CEIDARS). CEIDARS provides a foundation for the development of a more refined (hourly, grid-cell specific) set of emission inputs that are required by air quality models. The CEIDARS base year inventory is a primary input to the state s emission forecasting system, known as the California Emission Projection Analysis Model (CEPAM). CEPAM produces the projected emissions that are then gridded and serve as the emission input for the air quality models. The following sections of this document describe how base and future year emissions inventory estimates are prepared Inventory Coordination The California Air Resources Board convened the SIP Inventory Working Group (SIPIWG) to provide an opportunity and means for interested parties (CARB, districts, etc.) to discuss issues pertaining to the development and review of base year, future year, planning and gridded inventories to be used in SIP modeling. The group met every four to six weeks from March 2013 to May 2016 (ARB, 2016). Group participants included district staff from Bay Area, Butte, Eastern Kern, El Dorado, Feather River, Imperial, Northern Sierra, Placer, Sacramento, San Diego, San Joaquin, San Luis Obispo, South Coast, Ventura and Yolo-Solano. Additionally, CARB established the SIPIWG Spatial Surrogate Sub-committee, which focuses on improving input data to spatially disaggregate emissions at a more refined level needed for air quality modeling. Local air districts that participate include San 8

10 Joaquin Valley APCD, South Coast AQMD, Ventura County APCD and Sacramento Metropolitan AQMD. In addition to the two coordination groups described above, a great deal of work preceded this modeling effort through the Central California Air Quality Studies (CCAQS). CCAQS consisted of two studies: 1) the Central California Ozone Study (CCOS); and 2) the California Regional PM10 (particulate matter 10µ in diameter and smaller) /PM2.5 Air Quality Study (CRPAQS) Background California s emission inventory is an estimate of the amounts and types of pollutants emitted from thousands of industrial facilities, millions of motor vehicles and a myriad of emission sources such as consumer products and fireplaces. The development and maintenance of the emission inventory involves several agencies. This multi-agency effort includes: CARB, 35 local air pollution control and air quality management districts (Districts), regional transportation planning agencies (RTPAs), and the California Department of Transportation (Caltrans). CARB is responsible for the compilation of the final statewide emission inventory, and for maintaining this information in CEIDARS. In addition to the statewide emission inventory, emissions from northern Mexico (Kwong, 2017) are also incorporated in the final emission inventory used for modeling. The final emission inventory reflects the best information available at the time. The basic principle for estimating county-wide regulatory emissions is to multiply an estimated, per-unit emission factor by an estimate of typical usage or activity. For example, on-road motor vehicle emission factors are estimated for a specific vehicle type and applied to all applicable vehicles. The estimates are based on dynamometer tests of a small sample for a vehicle type. The activity for any given vehicle type is based on an estimate of typical driving patterns, number of vehicle starts, and typical miles driven. Assumptions are also made regarding typical usage; it is assumed that all vehicles of a certain vehicle type are driven under similar conditions in each region of the state. 9

11 Developing emission estimates for stationary sources involves the use of per unit emission factors and activity levels. Under ideal conditions, facility-specific emission factors are determined from emission tests for a particular process at a facility. A continuous emission monitoring system (CEMS) can also be used to determine a gas or particulate matter concentration or emission rate (U.S. EPA, 2016). More commonly, a generic emission factor is developed by averaging the results of emission tests from similar processes at several different facilities. This generic factor is then used to estimate emissions from similar types of processes when a facility-specific emission factor is not available. Activity levels from stationary sources can be derived from the amount of product produced, solvent used, or fuel used. The district reported and CARB estimated emissions totals are stored in the CEIDARS database for any given pollutant. Both criteria and toxic air pollutant emission inventories are stored in this complex database. These are typically annual average emissions for each county, air basin, and district. Modeling inventories for reactive organic gases (ROG) are estimated from total organic gases (TOG). Similarly, the modeling inventories for PM10 and PM2.5 are estimated from total particulate matter (PM). Details about chemical and size resolved speciation of emissions for modeling can be found in Section 0. Additional information on CARB emission inventories can be found at Inventory Years The emission inventory scenarios used for air quality modeling must be consistent with U.S. EPA s Modeling guidance (U.S. EPA, 2014). Since changes in the emissions inventory can affect the calculation of the relative response factors (RRFs) used to project air quality to future years, the terms used in the preparation of the emission inventory scenarios must be clearly defined. In this document, the following inventory definitions will be used: Base Case Modeling Inventory (2013) Base case modeling is intended to evaluate model performance and demonstrate confidence in the modeling system used for the modeled attainment test. The base 10

12 case modeling inventory is not used as part of the modeled attainment test itself. Model performance is assessed relative to how well model-simulated concentrations match actual measured concentrations. The modeling inputs are developed to represent (as best as possible) actual, day-specific conditions. Therefore, the base case modeling inventory for 2013 includes day-specific emissions for certain sectors. This includes, for instance, available day-specific activities and emission adjustments. Actual district reported point source emissions were gathered for the year 2012 and forecasted to The year 2013 was selected to coincide with the year selected for baseline design values (described below). The U.S. EPA modeling guidance states that once the model has been shown to perform adequately, the use of day-specific emissions is no longer needed. In preparation for SIP development, both CARB and the local air districts began a comprehensive review and update of the emission inventory several years ago resulting in a comprehensive emissions inventory for Reference Year Modeling Inventory (2013) The reference year inventory is intended to be a representation of emission patterns occurring through the baseline design value period and the emission patterns expected in the future year. U.S. EPA modeling guidance describes the reference year modeling inventory as a common starting point that represents average or typical conditions that are consistent with the baseline design value period. U.S. EPA guidance also states using a typical or average reference year inventory provides an appropriate platform for comparisons between baseline and future years. The 2013 reference year inventory represents typical average conditions and emission patterns through the 2013 design value period. This reference emissions inventory is not developed to capture day-specific emission characteristics. However, this reference inventory includes temperature, relative humidity and solar insolation effects, for Future Year Modeling Inventory (2020/2024/2025) Future year modeling inventories, along with the reference year modeling inventory, are used in the model-derived RRF calculation. Projected inventory years were chosen to address the following standards: 11

13 2020 is the modeled attainment year for the 24-hour 1997 PM2.5 standard of 65µg/m 3 and the annual 1997 PM2.5 standard of 15µg/m is the modeled attainment year for the 24-hour 2006 PM2.5 standard of 35µg/m is the modeled attainment year for the annual 2012 PM2.5 standard of 12 µg/m 3. Each of these years reflects the date by which attainment can be achieved as expeditiously as practicable for the relevant PM2.5 standard. These inventories maintain the typical, average patterns of the 2013 reference year modeling inventory. The 2020, 2024 or 2025 inventory will include temperature, relative humidity, and solar insolation effects from reference year (2013) meteorology. Future year point and area source emissions are projected from the 2012 baseline emissions used in the 2013 reference year modeling inventory. Additionally, future year on-road emission inventories are used, as projected by EMFAC. The application of control measure reduction factors is discussed in section Spatial Extent of Emission Inventories The emissions model-ready files that are prepared for use as an input for the air quality model conform to the definition and extent of the grids shown in Error! Reference source not found.. 12

14 Figure 1 Spatial coverage and parameter summary of modeling domains The domain uses a Lambert projection and assumes a spherical Earth. The emissions inventory grid uses a Lambert Conical Projection with two parallels. The parallels are at 30 and 60 N latitude, with a central meridian at W longitude. The coordinate system origin is offset to 37 N latitude. The emissions inventory uses a grid with a spatial resolution of 4 km x 4 km. The state modeling domain extends entirely over California and 100 nautical miles west over the Pacific Ocean. A smaller 4km x 4km subdomain is used for the San Joaquin Valley. The specifications for the statewide and San Joaquin Valley domains are summarized in Error! Reference source not found.. 13

15 Table 1 Modeling domain parameters Parameter Statewide domain San Joaquin Valley Subdomain Map Projection Lambert Conformal Conic Lambert Conformal Conic Datum None (Clarke 1866 None (Clarke 1866 spheroid) spheroid) 1st Standard Parallel 30.0 N 30.0 N 2nd Standard Parallel 60.0 N 60.0 N Central Meridian W W Latitude of projection origin 37.0 N 37.0 N COORDINATE SYSTEM Units Meters Meters Semi-major axis 6370 km 6370 km Semi-minor axis 6370 km 6370 km DEFINITION OF GRID Grid size 4km x 4km 4km x 4km Number of cells 321 x 291 cells 87 x 103 cells Lambert origin (-684,000 m, -564,000 m) (-108,000 m, -256,000 m) Geographic center Lat and 37.0 Lon Lat and 37.0 Lon 14

16 2. Estimation of Base Year Modeling Inventory As mentioned in Section 1.3, base case modeling is intended to demonstrate confidence in the modeling system used for the modeled attainment test. The following sections describe the temporal and spatial distribution of emissions and how each of the sectors within the modeling inventories are prepared Terminology The terms point sources and area sources are often confused. Traditionally, these terms have had different meanings to the developers of emissions inventories and the developers of modeling inventories. Error! Reference source not found. summarizes the difference in the terms. Both sets of terms are used in this document. In modeling terminology, point sources traditionally refer to elevated emission sources that exit from a stack and have an associated plume rise. While the current inventory includes emissions from stacks, all emission sources reported by the SJVAPCD (San Joaquin Valley Air Pollution Control District) associated with a facility are treated as potential elevated sources. The emissions processor calculates plume rise if appropriate; nonelevated sources are treated as ground-level sources. Examples of non-elevated emissions sources include gas dispensing facilities and storage piles. Area sources refers collectively to area-wide sources, stationary-aggregated sources, and other mobile sources (including aircraft, trains, ships, and all off-road vehicles and equipment). That is, area sources are low-level sources from a modeling perspective. 15

17 Table 2 Inventory terms for emission source types Modeling Term Emission Inventory Term Examples Point Stationary Point Facilities Stacks at Individual Facilities Area Area Off-Road Mobile Area-wide Construction Equipment, Farm Equipment, Trains, Recreational Boats Residential Fuel Combustion, Livestock Waste, Consumer Products, Architectural Coatings Area Stationary - Aggregated Industrial Fuel Use On-Road Motor Vehicles On-Road Mobile Cars and Trucks Biogenic Biogenic Trees The following sections describe in more detail the temporal, spatial and chemical disaggregation of the emissions inventory for point sources and area sources Temporal Distribution of Emissions The emissions are temporally resolved by month, week, day and hour to more accurately gauge model performance and ultimately to better assess the influence of control measures on attainment. This section covers the temporal distributions of the point, area, and off-road mobile sources. The temporal distribution of the emissions from on-road, biogenic and ocean-going vessel (OGV) are discussed in Sections 3.4, 3.5 and 3.6. The temporal distribution of residential wood combustion (RWC) and agricultural ammonia sectors are described in Section and Section 3.7.6, respectively. Temporal data are stored in CARB s emission inventory database. Each local air district assigns temporal data for all processes at each facility in their district to represent when emissions at each process occur. For example, emissions from degreasing may operate differently than a boiler. CARB or district staff also assign temporal data for each area source category by county/air basin/district. 16

18 Monthly Variation Emissions are adjusted temporally to represent variations by month. Some emission sources operate the same throughout a year. For example, a process heater at a refinery or a line haul locomotive likely operates the same month to month. Other emission categories, such as a tomato processing plant or use of recreational boats, vary significantly by season. CARB s emission inventory database stores the relative monthly fractional activity for each process, the sum of which is 100. Using an example of emission sources that typically operate the same over each season, emissions from refinery heaters and line haul locomotives would have a monthly fraction (throughput) of 8.33 for each month (calculated as 100/12 = 8.33). This is considered a flat monthly profile. To apply monthly variations to create a gridded inventory, the annual average day s emissions (yearly emissions divided by 365) is multiplied by the typical monthly throughput. For example, a typical monthly throughput in July for recreational boats of 15 results in about 1.8 times higher (15 / 8.33 = 1.8) emissions than a day in a month with a flat monthly profile Weekly Variation Emissions are adjusted temporally to represent variations by day of week. Some operations are the same over a week, such as a utility boiler or a landfill. Many businesses operate only 5 days per week. Other emissions sources are similar on weekdays, but may operate differently on weekend days, such as architectural coatings or off-road motorcycles. To accommodate variations in days of the week, each process or emission category is assigned a days-per-week code or DPWK. Error! Reference source not found. shows the current DPWK codes and Error! Reference source not found. in Appendix D shows additional DPWK codes used for agricultural-related emissions. 17

19 Table 3 Day of week variation factors Code WEEKLY CYCLE CODE DESCRIPTION M T W TH F S S 1 One day per week Two days per week Three days per week Four days per week Five days per week - Uniform activity on week days; non on Saturday and Sunday Six days per week - Uniform activity on week days; non on Saturday and Sunday Seven days per week Uniform activity every day Of the week Uniform activity on Saturday and Sunday; no activity the remainder of the week Uniform activity on Saturday and Sunday; half as much activity on week days Uniform activity on week days; reduced activity on weekends Uniform activity on week days; reduced activity on weekends (For onroad motor vehicles) Uniform activity on week days; half as much activity on Saturday. Little activity on Sunday Uniform activity on week days; one third as much on Saturday; little on Sunday Uniform activity on week days; little activity on Saturday; no activity on Sunday Uniform activity on week days; half as much activity on weekends Uniform activity on week days; five times as much activity on weekends Uniform activity on Monday through Thursday; increased activity on Friday, Saturday, Sunday Daily Variation Emissions are adjusted temporally to represent variations by hour of day. Many emission sources occur 24 hours per day, such as livestock waste or a sewage treatment plant whereas many businesses operate 8 hours per day. Other emissions sources vary significantly over a day, such as residential space heating or pesticide application. Each process or emission category is assigned an hours-per-day (HPDY) code. Error! Reference source not found. displays the daily variation factors or current HPDY codes. These codes are mostly current except for Code 33 which changed in response to RWC temporal allocation methods (see section 3.7.5). Specifically, the morning-evening peak pattern is replaced with an evening-only profile up to midnight. Table 18 in Appendix D shows additional DPWK codes used for agricultural-related emissions. 18

20 Table 4 Daily variation factors Code CODE DESCRIPTION HOUR PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY - UNIFORM ACTIVITY FROM 8 A.M. TO 4 P.M. (NORMAL WORKING SHIFT) HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY - UNIFORM ACTIVITY FROM 8 A.M. TO MIDNIGHT (2 WORKING SHIFTS) HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY HOURS PER DAY - UNIFORM ACTIVITY DURING THE DAY MAJOR ACTIVITY 5-9 P.M., AVERAGE DURING DAY, MINIMAL IN EARLY A.M.(GAS STATIONS) MAX ACTIVITY 7-9 A.M. & 7-11 P.M.,AVERAGE DURING DAY, LOW AT NIGHT (RESIDENTIAL FUEL COMBUSTION) ACTIVITY 1 TO 9 A.M.; NO ACTIVITY REMAINDER OF DAY (i.e. ORCHARD HEATERS) MAX ACTIVITY 7 A.M. TO 1 A.M., REMAINDER IS LOW (i.e. COMMERCIAL AIRCRAFT) ACTIVITY DURING DAYLIGHT HOURS; LESS CHANCE IN EARLY MORNING AND LATE EVENING ACTIVITY DURING MEAL TIME HOURS (i.e. RESIDENTIAL COOKING) PEAK ACTIVITY AT 7 A.M. & 4 P.M.; AVERAGE DURING DAY (ON-ROAD MOTOR VEHICLES) ACTIVITY FROM 6 A.M. TO 12 P.M. (PETROLEUM DRY CLEANING) MAJOR ACTIVITY FROM 6 A.M.-12 P.M., LESS FROM 12-7 P.M. (PESTICIDES) ACTIVITY FROM 7 A.M. TO 12 P.M. (AGRICULTURAL AIRCRAFT) UNIFORM ACTIVITY FROM 7 A.M. TO 9 P.M. (DAYTIME BIOGENICS) UNIFORM ACTIVITY FROM 9 P.M. TO 7 A.M. (NIGHTIME BIOGENICS) MAX ACTIVITY 8 A.M. TO 5 P.M, MINIMAL AT NIGHT & EARLY MORNING(CAN&COIL/METAL PARTS COATINGS) MAX ACTIVITY 7 A.M. TO 2 P.M., MINIMAL AT EVENING AND MORNING HOURS (CONSTRUCTION EQUIPMENT ON HOT MAX ACTIVITY 7 A.M. TO NOON.;REDUCED ACTIVITY NOON TO 6 P.M. (AUTO REFINISHING) MAXIMUM ACTIVITY FROM 7:00 AM TO 3:00 PM; REDUCED ACTIVITY FROM 3:00 TO 6:00 PM.(CONSTRUCTION MAXIMUM ACTIVITY FROM NOON TO 7:00 PM; REDUCED ACTIVITY EVENING AND MORNING HOURS (RECREATIONAL MAX ACTIVITY 9 AM TO 3 PM; HALF THE ACTIVITY REMAINING HOURS (WASTE FROM DAIRY CATTLE) ACTIVITY FROM 10 AM TO 9 PM RISING TO PEAK AT 3; NO ACTIVITY REMAINDER OF DAY (WASTE FROM POULTRY) ACTIVITY FROM 9 AM TO 12 AM RISING TO PEAK AT 3; MINIMUM ACTIVITY REMAINDER OF DAY (WASTE FROM SWINE) MAJOR ACTIVITY FROM 11AM TO 6PM; REDUCED OTHER HOURS (EVAP-COASTAL COUNTIES) MAJOR ACTIVITY FROM 11AM TO 6PM; REDUCED OTHER HOURS (EVAP-NON-COASTAL COUNTIES)

21 2.3. Spatial Allocation Once the base case, reference or future year inventories are developed, the next step of modeling inventory development is to spatially allocate the emissions. Air quality models attempt to replicate the physical (e.g. transport) and chemical processes that occur in the atmosphere within a modeling domain. Therefore, it is important that the physical location of emissions be specified as accurately as possible. Ideally, the actual location of all emissions would be known exactly. In reality, however, some categories of emissions would be virtually impossible to determine for example, the actual amount and location of consumer products (e.g. deodorant) used every day. To the extent possible, the spatial allocation of emissions in a modeling inventory approximates as closely as possible the actual location of emissions. Spatial allocation is typically accomplished by using spatial surrogates. These spatial surrogates are processed into spatial allocation factors in order to geographically distribute county-wide area source emissions to individual grid cells. Spatial surrogates are developed based on demographic, land cover and other data that exhibit patterns that vary geographically. Sonoma Technology, Inc. (STI) (Funk, et al., 2001) under CCOS contract originally developed many of the spatial surrogates by creating a base year (2000) and various future year surrogate inventories. STI updated the underlying spatial data and developed new surrogates (Reid, et al., 2006), completing the project in CARB and districts have since continued to update and improve many of the spatial surrogates, adding new ones as more data become available. Three basic types of surrogate data were used to develop the original spatial allocation factors: land use and land cover; facility location; and demographic and socioeconomic data. Land use and land cover data are associated with specific land uses, such as agricultural harvesting or recreational boats. Facility locations are used for sources such as gas stations and dry cleaners. Demographic and socioeconomic data, such as population and housing, are associated with residential, industrial, and commercial activity (e.g. residential fuel combustion). To develop spatial allocation factors of high quality and resolution, local socioeconomic and demographic data were used when available for developing base case, baseline and future year inventories. These data 20

22 were available from local Metropolitan Planning Organizations (MPO) or Regional Transportation Planning Agencies (RTPA), where they are used as inputs for travel demand models. In rural regions for which local data were not available, data from Caltrans Statewide Transportation Model were used. Since 2008, CARB and district staffs have continued to search for more recent or improved sources of data, since the underlying data used by STI were prior to the recession. CARB and district staffs have updated many of the spatial surrogates and added many new ones. Updates to land use categories were made using the National Land Cover Database 2011 (Homer, et al., 2015). Many surrogates were updated using the locations from Dun & Bradstreet s Market Insight Database (Dun and Bradstreet, 2015). The types of sources were defined by SIC. Fourteen new surrogates were developed for industrialrelated sources using SIC and whether manufacturing occurred at the facility. U.S. Census American Community Survey (FactFinder, 2011) data by census block were used to update residential fuel use. Sierra Research developed nine new surrogates related to agricultural activities (Anderson, et al., 2012), some of which incorporated crop-specific factors. Seven new surrogates were developed using vessel traffic data, or Automatic Identification System (AIS) data, collected by the U.S. Coast Guard. A new surrogate was created to represent the location of construction equipment. The distribution is a combination of two sets of data: 90% change in imperviousness between 2006 and 2011 from NLCD 2011 and 10% road network. Impervious surfaces are mainly artificial structures such as pavements (roads, sidewalks, driveways and parking lots) that are covered by materials impenetrable to a satellite such as asphalt, concrete, brick, stone and rooftops. A new surrogate was compiled to distribute emissions from transport refrigeration units (TRU) from three sources: 65% distribution centers, 34% road network and 1% grocery stores / food processing facilities. Information on distribution centers were retrieved from ARBER, the CARB Equipment 21

23 Registration database for the Transport Refrigeration Unit (TRU) ATCM and the Drayage Truck Regulation. Development of a wilderness mask for application of population based spatial surrogates (see below paragraphs) Utilization of Digital Map Products California state-wide parcel database to develop a new fireplace surrogate (see section 2.3.7) In all, 99 unique surrogates are available for use. A summary of the spatial surrogates for which spatial allocation factors were developed is shown below in Error! Reference source not found.. Table 5 Spatial Surrogates Surrogate Name AEROSPACE Airports All_PavedRds AutobodyShops CAFO CANCOIL Cemeteries Comm_Airports COMPOST CONSTRUCTION_EQUIP Devplnd_HiDensity Devplnd_LoDensity DREDGE Drycleaners DryLakeBeds Elev5000ft Employ_Roads FABRIC FERRIES FISHING_COMM Forestland Fugitive_Dust GAS_DISTRIBUTION GAS_SEEP GasStations GASWELL GolfCourses HE_Sqft Hospitals Housing Housing_Autobody Housing_Com_Emp Housing_Restaurants Surrogate Name INDUSTRIAL Industrial_Emp InlandShippingLanes Irr_Cropland Lakes_Coastline Surrogate Definition Spatial distribution of businesses involved in aerospace Spatial locations of all airports Spatial distribution of road network (all paved roads) Locations of autobody repair and refinishing shops Spatial distribution of concentrated animal feeding operations Spatial distribution of businesses involved in can and coil operations Spatial locations of cemeteries Spatial locations of commercial airports Spatial distribution of composting Spatial distribution of where construction equipment is used Spatial distribution of developed land - low density, medium density and high density Spatial distribution of developed land - open space (lowest density) Locations of dredging Locations of dry cleaning facilities Locations of dry lake beds Topological contours areas above 5000 feet Spatial distribution of total employment and road density (all paved roads) Spatial distribution of businesses involved in fabric manufacturing Locations of ferry ports and routes Locations of commercial fishing Spatial distribution of forest land Spatial distribution of barren land Location of gas pipelines Location of natural-occurring gas seeps Locations of gasoline service stations Locations of gas wells Spatial locations of golf courses Computed surrogate based on housing and employment (est. ft2 / person) Spatial locations of hospitals Spatial distribution of total housing Spatial distribution of housing and autobody refinishing shops Spatial distribution of total housing and commercial employment Spatial distribution of total housing and restaurants/bakeries Surrogate Definition Spatial distribution of industrial businesses where manufacturing occurs (SIC<4000) Spatial distribution of industrial employment Spatial distribution of major shipping lanes within bays and inland areas Spatial location of agricultural cropland Locations of lakes, reservoirs, and coastline 22

24 Surrogate Name LAKES_RIVERS_RECBOAT LANDFILLS LANDPREP LINEHAUL LiveStock MARINE METALFURN METALPARTS Metrolink_Lines MILITARY_AIRCRAFT MILITARY_SHIPS MILITARY_TACTICAL MiltaryBases NON_PASTURE_AG NonIrr_Pastureland NonRes_Chg OCEAN_RECBOAT OIL_SEEP OILWELL OTHERCOAT PAPER PASTURE PEST_ME_BR PEST_NO_ME_BR PLASTIC Pop_ComEmp_Hos Population Ports POTWs PrimaryRoads PRINT Raillines RailYards Rds_HE RefinieriesTankFarms Res_NonRes_Chg ResGasHeating Residential_Chg ResLPGHeat ResNonResChg_IndEmp ResOilHeat Restaurants ResWoodHeating FIREPLACES Surrogate Name SandandGravelMines Schools SecondaryPavedRds SEMICONDUCT Ser_ComEmp_Sch_GolfC_Cem Service_Com_Emp Shiplanes SILAGE SingleHousingUnits TRU TUG_TOW UnpavedRds Wineries WOOD WOODFURN Surrogate Definition Locations of lakes, rivers and reservoirs where recreational boats are used Locations of landfills Spatial distribution of dust from land preparation operations (e.g. tilling) Spatial distribution of Class I rail network Spatial distribution of cattle ranches, feedlots, dairies, and poultry farms Spatial distribution of businesses involved in marine Spatial distribution of businesses involved in metal furniture Spatial distribution of businesses involved in metal parts and products Spatial distribution of metrolink network Locations of landing strips on military bases Locations of military ship activity Military bases where tactical equipment are used Locations of military bases Spatial distribution of farmland Spatial location of pasture land Computed surrogate based on spatial distribution of non-residential areas Locations of recreational boat activity that can occur on the ocean and SF Bay Location of naturally-occurring oil seeps Locations of oil wells (both onshore and offshore) Spatial distribution of businesses with SIC<4000 not included in another category Spatial distribution of businesses involved in paper Spatial distribution of grazing land Spatial distribution of methyl bromide pesticides Spatial distribution of non-methyl bromide pesticides Spatial distribution of businesses involved in plastic Spatial distribution of hospitals, population and commercial employment Spatial distribution of population Locations of shipping ports Coordinate locations of publicly owned treatment works (POTWs) Spatial distribution of road network (primary roads) Spatial distribution of print businesses Spatial distribution of railroad network Locations of rail yards Calculated surrogate based on road densities and housing/employment (est. ft2 / person) Coordinate locations of refineries and tank farms Computed surrogate based on spatial distribution of residential and non-residential areas Spatial distribution of homes using gas supplied by a utility as primary source of heating Computed surrogate based on spatial distribution of residential areas Spatial distribution of homes using gas (bottled, tank or LP) as primary source of heating Spatial distribution of industrial employment and residential/non-residential change Spatial distribution of homes using fuel oil or kerosene as primary source of heating Locations of restaurants Spatial distribution of homes using wood as primary source of heating Spatial distribution of residential wood heating actually being used by RWC woodstoves and fireplaces Surrogate Definition Locations of sand/gravel excavation and mining Spatial locations of schools Spatial distribution of road network (secondary roads) Spatial distribution of businesses involved in semiconductors Spatial distribution of service and commercial employment, schools, cemeteries,golf courses Spatial distribution of service and commercial employment Spatial distribution of major shipping lanes Spatial distribution of silage operations Spatial distribution of single dwelling units Spatial distribution of transport refrigeration units Spatial distribution of tug and tow boats Spatial distribution of road network (unpaved roads) Locations of wineries Spatial distribution of businesses using wood Spatial distribution of businesses involved in wood furniture 23

25 Recent updates to the spatial surrogates include the creation of a wilderness mask. A wilderness mask was developed by CARB staff to incorporate land area in certain grid cells that had reported population from the census block but in reality are remote or desolate wilderness. Error! Reference source not found. illustrates a wilderness surrogate developed based on compiled data from the USFS (United States Forest Service), National Park Service, BLM (Bureau of Land Management) and state park systems. Wilderness is defined as an area of undeveloped land without permanent improvements or human habitation (Funk, et al., 2001). Figure 2 Wilderness mask on ca-4km state domain The wilderness mask was applied to approximately 20 surrogates to remove fractions of disaggregated emissions in areas where no humans live. Error! Reference source not found.3 illustrates how the wilderness mask removes county fractions in certain 24

26 areas in the population surrogate while table 6 describes all surrogates in which the mask was applied. (a) (b) Figure 3 Example of population surrogate before (a) and after (b) wilderness mask application 25

27 Table 6 List of Surrogates in which the wilderness mask was applied Description Number / Shape File Name 180 Employ Roads Spatial distribution of total employment and road density (all paved roads) 230 HE_Sqft Computed surrogate based on housing and employment (est. ft2 / person) 250 Housing Spatial distribution of total housing 260 Housing Autobody Spatial distribution of housing and auto-body refinishing shops 270 Housing_Com_Emp Spatial distribution of total housing and commercial employment 280 Housing_Restaurants Spatial distribution of total housing and restaurants/bakeries 300 Industrial_Emp Spatial distribution of industrial employment 440 Population Population 450 Pop_ComEmp_Hos Spatial distribution of hospitals, population and commercial employment 530 ResGasHeating Spatial distribution of homes using gas supplied by a utility as primary source of heating Residential Change Computed surrogate based on spatial distribution of residential areas Res Non Res Change Industrial Employment Spatial distribution of industrial employment and residential/nonresidential change Res Wood Heating Spatial distribution of homes using wood as primary source of heating Res Oil Heating Spatial distribution of homes using fuel oil or kerosene as primary source of heating Res LPG Heating Spatial distribution of homes using gas (bottled, tank or LP) as primary source of heating Fireplaces Spatial distribution of residential wood heating actually being used by RWC woodstoves and fireplaces Res Non Res Change Computed surrogate based on spatial distribution of residential and nonresidential areas Service Com Employment Spatial distribution of service and commercial employment 26

28 650 - Ser Com Emp Schools Golf Course Cemeteries Spatial distribution of service and commercial employment, schools, cemeteries, golf courses Developed Land High Density Spatial distribution of developed land - low density, medium density and high density The following sections describe in more detail the type of spatial disaggregation used for each sector of the emissions inventory Spatial Allocation of Area Sources Each area source category is assigned a spatial surrogate that is used to allocate emissions to a grid cell in CARB s 4km statewide modeling domain. Examples of surrogates include population, land use, and other data with known geographic distributions for allocating emissions to grid cells, as described above Spatial Allocation of Point Sources Each point source is allocated to grid cells using the latitude and longitude reported for each stack. If there are no stack latitude and longitude, the facility coordinates are used. There are two types of point sources: elevated and non-elevated sources. Vertical distribution of elevated sources is allocated using the plume rise algorithm in the emissions processor, Sparse Matrix Operator Kernel Emissions (SMOKE) (see Section 3.3), while non-elevated are allocated to the first layer. Most stationary point sources with existing stacks are regarded as elevated sources. Those without physical stacks that provide only latitude/longitude, such as airports or landfills, are considered non-elevated Spatial Allocation of Wildfires, Prescribed Burns and Wildland Fire Use Emissions from these sources are event and location-based. A fire event can last a few hours or span multiple days. Each fire is spatially allocated to grid cells using the extent of each fire event while the temporal distribution also reflects the actual duration of the fire. The spatial information to allocate the fire emissions comes from a statewide interagency fire perimeters geodatabase maintained by the Fire and Resource 27

29 Assessment Program (FRAP) of the California Department of Forestry and Fire Protection (CALFIRE). More details on the methodology and estimation of the wildfire emissions can be found in Section Spatial Allocation of Ocean going vessels (OGV) Ship emissions are allocated to the grids corresponding to the vessel traffic lanes in CARB s OGV model (ARB-PTSD, 2011) These traffic lanes were estimated from three different sources: 1.) National Waterway Network, 2.) The Ship Traffic and 3.) Energy and Environment Model Automated instrumentation system (AIS) telemetry data collected in Spatial Allocation of On-road Motor Vehicles The spatial allocation of on-road motor vehicles is based on DTIM as described in Section Spatial Allocation of Biogenic Emissions As described in Section 3.5, gridded biogenic emissions are derived using the Model of Emissions of Gases and Aerosols from Nature (MEGAN). MEGAN utilizes gridded emission factors and plant functional type data, adjusted by local meteorological conditions and satellite derived leaf area data, to estimate hourly biogenic emissions within each grid cell of the modeling domain. More details about MEGAN can be found at Spatial Allocation of Residential Wood Combustion Emissions A parcel database developed by Digital Map Products was utilized to create a new spatial surrogate for residential wood combustion. The CA-statewide parcel data included a vast amount of data regarding property information. A specific true/false tag for fireplaces was provided for each parcel of land. The database was filtered for properties that had a fireplace tag set to true and then manually filtered for inconsistencies based on land type and reported tag (example: non-residential and vacant agricultural parcels that were reported to have a fireplace tag were removed based on conversation with CARB staff). 28

30 For some counties, the data were extremely limited. This may be the result of building restrictions and permit issues for fireplaces in new homes. Therefore, for the counties in which the parcel data seemed unreasonable, the original RWC surrogate (570) was used as default. The counties where the original default surrogate was applied are San Diego, Los Angeles, Mendocino, Ventura, Santa Barbara, San Louis Obispo, Imperial, Contra Costa, Alameda, and Del Norte. Error! Reference source not found. shows the new fireplace surrogate, # 573 and illustrates the spatial distribution of residential wood heating being used by woodstoves and fireplaces. Figure 4 New spatial surrogate of fireplaces based on housing population (2017) 29

31 Residential wood curtailment (i.e. no-burn days) was applied after spatial allocation for each district. Figure 5 illustrates where emissions are reduced due to residential wood curtailment programs in three air districts, and more description regarding curtailment methods are discussed in section Figure 5 Map of residential wood curtailment areas In the San Joaquin Valley, a reduction in emissions due to curtailment was only applied to areas where natural gas service is available (e.g. provided by a municipality) as reflected in Rule 4901 (October 2008 version of the rule). In the South Coast, emission reductions due to curtailment were applied to locations below 3000 feet, as stated in Rule 445. For Sacramento, curtailment was applied to all of Sacramento County per Rule

32 2.4 Speciation Profiles CARB s emission inventory lists the amount of pollutants discharged into the atmosphere by source in a certain geographical area during a given time period. It currently contains estimates for CO, NH3, NOx, SOx, total organic gases (TOG) and PM. CO and NH3 each are single species; NOx emissions are composed of NO, NO2 and HONO; and SOx emissions are composed of SO2 and SO3. TOG and PM potentially contain over hundreds of different chemical species, and speciation is the process of disaggregating those inventory pollutants into individual chemical species components or groups of species. CARB maintains and updates such species profiles for organic gases (OG) and PM for a variety of source categories. Photochemical models simulate the physical and chemical processes in the lower atmosphere, and include all emissions of the important classes of chemicals involved in photochemistry as well as less reactive compounds that are of concern from a health or visibility standpoint. Organic gases emitted to the atmosphere are referred to as Total Organic Gas or TOG. TOG includes all organic compounds that can become airborne (through evaporation, sublimation, as aerosols, etc.), excluding carbon monoxide, carbon dioxide, carbonic acid, metallic carbides or carbonates and ammonium carbonate. TOG emissions reported in the CARB s emission inventory are the basis for deriving the Reactive Organic Gas (ROG) emission components, which are also reported in the inventory. ROG is defined as TOG minus CARB s exempt compounds (e.g., methane, ethane, various chlorinated fluorocarbons, acetone, perchloroethylene, volatile methyl siloxanes, etc.). ROG is nearly identical to U.S. EPA s Volatile Organic Compounds (VOC), which is based on EPA s exempt list. For all practical purposes, use of the terms ROG and VOC are interchangeable. The OG speciation profiles are applied to estimate the amounts of various organic compounds that make up TOG emissions. A speciation profile contains a list of organic compounds and the weight fraction that each compound comprises of the TOG emissions from a particular source type. In addition to the chemical name for each chemical constituent, the file also shows the 5-digit CARB internal identification 31

33 chemical code The speciation profiles are applied to TOG to develop both the photochemical model inputs and the emission inventory for ROG. It should be noted that districts are allowed to report their own reactive fraction of TOG that is used to calculate ROG rather than use the information from the assigned organic gas speciation profiles. These district-reported fractions are not used in developing modeling inventories because the information needed to calculate the amount of each organic compound is not available. The PM emissions are size fractionated by using PM size distribution profiles, which contain the total weight fraction for PM2.5 and PM10 out of total PM. The fine and coarse PM chemical compositions are characterized by applying the PM chemical speciation profiles for each source type, which contain the weight fractions of each chemical species for PM2.5, PM10, and total PM. PM chemical speciation profiles may also vary for different PM size fractions even for the same emission source. PM size profiles and speciation profiles are typically generated based on source testing data. In most previous source testing studies aimed at determining PM chemical composition, filterbased sampling techniques were used to collect PM samples for chemical analyses. The original OG profiles and PM profiles are available for download from CARB s web site at: Based on these original profiles, a model-ready speciation file, gspro, was generated for a specific chemical mechanism (for example, SAPRC07) to separate aggregated inventory pollutant emission totals into emissions of model species required by the AQM (air quality model). Each process or product category is keyed to one of the OG profiles and one of the PM profiles. Also available for download from CARB s web site (see link in previous paragraph) is a cross-reference file that indicates which OG profile and PM profile are assigned to each category in the inventory. The inventory source categories are represented by an 8-digit source classification code (SCC) for point sources, or a 14- digit emission inventory code (EIC) for area and mobile sources. Some of the organic gas profiles and PM profiles related to motor vehicles, ocean going vessels, and fuel evaporative sources vary by the inventory year of interest, due to changes in fuel 32

34 composition, vehicle fleet composition and emissions control devices such as diesel particulate filters (DPFs).. Details can be found in CARB s documentation of speciation profile development. Mapping of each category to OG and PM profiles is summarized in rogpm and gsref files. Research studies are conducted regularly to improve CARB s speciation profiles. These profiles support ozone and PM modeling studies but are also designed to be used for regional toxics modeling. Other health or welfare related modeling studies where the compounds of interest cannot always be anticipated make use of these profiles. Therefore, speciation profiles need to be as complete and accurate as possible. CARB has an ongoing effort to update speciation profiles as data become available, such as through testing of emission sources or surveys of product formulations. New speciation data generally undergo technical and peer review, and updating of the profiles is coordinated with users of the data. The recent addition to CARB s speciation profiles include (ARB, 2017) : (1) Organic gas profile Consumer products Architectural coating Gasoline fuel and headspace vapor Gasoline vehicle hot soak and diurnal evaporation Gasoline vehicle start and running exhaust Silage Aircraft exhaust Compressed Natural Gas (CNG) bus running exhaust (2) PM profile Gasoline vehicle exhaust On-road diesel exhaust Off-road diesel exhaust Ocean going vessel exhaust Aircraft exhaust Concrete batching Commercial cooking Residential fuel combustion-natural gas Coating/painting 33

35 Cotton ginning Stationary combustion 34

36 3 Methodology for Developing Base Case, Baseline, and Future Projected Emissions Inventories As mentioned in Section 0, the base case and reference inventories include temperature, humidity and solar insolation effects for some emission categories; development of these data is described in Sections 3.1 and 3.2. The remaining sections of Chapter 3 detail how the base case and reference inventories were created for different sectors of the inventory such as point, area, on-road motor vehicles, biogenic, OGV and other day-specific sources. 3.1 Surface Temperature and Relative Humidity Fields The calculation of gridded emissions for some categories of the emissions inventory is dependent on various meteorological variables. As an example, biogenic emissions are sensitive to air temperatures and solar radiation while emissions from on-road mobile sources are sensitive to air temperature and relative humidity. Therefore, estimates of air temperature (T), relative humidity (RH), and solar radiation are needed for each grid cell in the modeling domain in order to take into account the effects of these meteorological variables. Gridded temperature and humidity fields are readily available from prognostic meteorological models such as the Weather Research and Forecasting (WRF) model ( which is used to prepare meteorological inputs for the air quality model. However, prognostic meteorological models can at times have difficulty capturing diurnal temperature extremes (Valade, 2009; Caldwell, 2009; Fovell, 2008). Since temperature and the corresponding relative humidity extremes can have an appreciable influence on some emissions categories, such as on-road mobile and biogenic sources, measurement based fields for these parameters are used in processing emissions. The CALMET ( diagnostic meteorological model is utilized to generate both the gridded temperature and relative humidity fields used in processing emissions. The principal steps involved in generating a gridded, surface-level temperature field using CALMET include the following: 35

37 1. Compute the relative weights of each surface observation station to each grid cell (the weight is inversely proportional to the distance between the surface observation station and grid cell center). 2. Adjust all surface temperatures to sea level. In this step, a lapse rate of o C/m is used (this lapse rate is based on private communication with Gary Moore of Earth Tech, Inc., Concord, MA). This lapse rate (=2.7 F/1000 feet) is based on observational data. 3. Use the weights to compute a spatially-averaged sea-level temperature for each grid cell. 4. Correct all sea-level temperatures back to 10 m height above ground level (i.e. the standard height of surface temperature measurements) using the lapse rate of o C/m again. 5. The current version of CALMET does not generate estimates of relative humidity. As a result, a post-processing program was used to produce gridded, hourly relative humidity estimates from observed relative humidity data. The major steps needed to generate gridded, surface-level relative humidity are described as follows: a. Calculate actual vapor pressure from observed relative humidity and temperature at all meteorological stations. The (Mc. Rae, 1980) method is used to calculate the saturated vapor pressure from temperature; b. Compute the relative weights of each surface observation station to each grid in question, exactly as done by CALMET to compute the temperature field; c. Use the weights from step 2 to compute a spatially-averaged estimate of actual vapor pressure in each grid cell; d. For each grid cell, calculate relative humidity from values for actual vapor pressure and temperature for the same grid cell. 3.2 Insolation Effects Insolation data was used in the estimation of the gridded emissions inventory and provided by the WRF meteorological fields as mentioned in Section

38 3.3 Estimation of Gridded Area and Point sources Emissions inventories that are temporally, chemically, and spatially resolved are needed as inputs for the photochemical air quality model. Point sources and area sources (area-wide, off-road mobile and aggregated stationary) are processed into emissions inventories for photochemical modeling using the SMOKE modeling system ( California-specific improvements to SMOKE were implemented under a CARB contract for version 4.0 of SMOKE (Baek, 2015). However, GenTpro, a pre-smoke utility program that modulates annual hourly temporal profiles based on modeled meteorology, cannot run in CARB SMOKE due to the fact that it does not recognize the COABDIS defined region code as an acceptable alphanumeric parameter. In 2018, CARB SMOKE was replaced with the CMASreleased SMOKEv4.0 (referred as Official SMOKE hereafter) and included changes to the GenTpro program that accepted the numeric CARB GAI region code. COABDIS based cross-reference files were subsequently changed to GAI in order to match the Official SMOKE format. Inputs for SMOKE are annual emissions totals from CEPAM and information for allocating to temporal, chemical, and spatial resolutions. Temporal inputs for SMOKE are screened for missing or invalid temporal codes as discussed in Section 4.1. Temporal allocation of emissions using SMOKE involves the disaggregation of annual emissions totals into monthly, day of week, and hour of day emissions totals. The temporal codes from Error! Reference source not found. and Error! Reference source not found. are reformatted into an input-ready format as explained in the SMOKE user s manual. Chemical speciation profiles, as described in Section 0, and emissions source cross-reference files used as inputs for SMOKE are developed by CARB staff. SMOKE uses the files for the chemical speciation of NOx, SOx, TOG and PM to produce the species needed by photochemical air quality models. Emissions for area sources are allocated to grid cells as stated by the modeling grid domain defined in Section 1.4. Emissions are spatially disaggregated by the use of spatial surrogates as described in Section 2.3. These spatial surrogates are converted to a SMOKE-ready format as described in the SMOKE user s manual. Emissions for 37

39 point sources are allocated to grid cells by SMOKE using the latitude and longitude coordinates reported for each stack. 3.4 Estimation of On-road Motor Vehicle Emissions The EMFAC emissions model is used by CARB to assess emissions from on-road vehicles including cars, trucks, and buses in California, and to support air quality planning efforts to meet the Federal Highway Administration's transportation planning requirements. EMFAC is designed to produce county-level, average-day estimates. As a result, these estimates must be disaggregated spatially and temporally into gridded, hourly estimates for air quality modeling. The general methodology used to disaggregate EMFAC emission estimates is a twostep approach. The first step uses the Direct Travel Impact Model (DTIM4) (Systems Applications Inc., 2001) to produce gridded, hourly emission estimates. The second step distributes EMFAC emissions according to the spatiotemporal output from DTIM. This methodology has been peer reviewed by the Institute of Transportation Studies at the University of California, Irvine, under CCOS contract 11-4CCOS. The spatiotemporal allocation of emissions from DTIM does not vary dramatically with small changes in meteorological data (T/RH), resulting in a negligible monthly variation of the spatial surrogate. However, differences in DTIM s winter versus summer spatiotemporal allocation are slightly appreciable. Therefore, different spatial surrogates are created for a winter and a summer day. At the time of the development of these inventories the most recent version of EMFAC that has been approved by USEPA for SIP and conformity purposes is EMFAC2014 (80 FR 77337). EMFAC2014 has three separate modules that are relevant for the preparation of the on-road emissions gridded inventory: one that estimates emissions, one that estimates emission rates, and one that estimates activity data. The emissions module runs for every county and every day of the modeled year using day-specific temperature and relative humidity. On a less granular level, the emissions rates module runs for every county for a summer day and a winter day. Lastly, the activity module 38

40 runs once to estimates vehicle miles traveled (VMT), number of vehicle trips, fuel consumption, and the number of vehicles in use General Methodology Mobile source emissions are sensitive to ambient temperature and humidity. Both EMFAC and DTIM account for meteorological effects using day-specific inputs. For EMFAC, hourly gridded temperature and humidity fields are averaged by county using a gridded VMT weighted average (i.e. weighted proportional to the VMT per grid cell in a county). DTIM accepts gridded, hourly data directly (CALMET formatted data). See Section 3.1 for more information on CALMET. EMFAC provides vehicle-class and fuel specific emissions estimates for: exhaust, evaporative, tire wear, and brake wear emissions. EMFAC also produces estimates of: VMT, number of vehicle trips, fuel consumption, and the number of vehicles in use. More information on EMFAC can be found at (ARB-MSEI, 2015). The vehicle activity is the most important input for spatiotemporal distribution of emissions. DTIM uses hourly vehicle miles traveled on each highway link and each of the vehicle trips in the modeling domain. The detailed vehicle activity data is obtained from CARB s Integrated Transportation Network (ITN) version 3 database. The overall processing of on-road emissions to create the gridded emissions inventory is shown in Error! Reference source not found.6. Activity data from the ITN (see Section3.4.2) is developed for the thirteen EMFAC 2007 vehicle types, but activity is split for gas and diesel, resulting in a total of 26 vehicle types as shown in the block diagram. The forecasted on-road modeling inventories are developed using the same methodology as the baseline year, where future year emissions are based on running EMFAC 2014 in Emissions Mode for the associated future year. 39

41 Daily CALMET data EMFAC Emissions Daily runs (365 runs) EMFAC Emission rates Winter/Summe ITN activity data for 26 vehicle types DTIM* hdd emissions winter/summer ldd,ldg,hdg emissions Every day (365 runs) Configuration files* DTIM MEDS ** Scaling Winter/Summer Gridded on-road file (pmeds)*** *DTIM is run four times - once for each of: ldg (light-duty gas), ldd (light-duty diesel), hdg (heavy-duty gas), hdd (heavy-duty diesel) **DTIM MEDS is a similar format to the MEDS format ***pmeds is a revised version of the MEDS format that has increased precision for the emission values Day of week factors by light/medium, light heavy, heavy heavy duty Figure 6 Block diagram for on-road processing 40

42 3.4.2 ITN Activity Data The ITN is a database which is populated with link-based and Traffic Analysis Zone (TAZ)-based travel activity from travel demand models provided by different metropolitan planning organizations (MPOs), California Department of Transportation (Caltrans) and other California regional transportation planning agencies. The vintage and types of data used in the current version of the ITN are shown in Error! Reference source not found.7. Different types of quality control parameters like vehicle mix, hourly distributions and post-mile coverage are obtained from default EMFAC and Caltrans databases. After these various pieces of data are imported to the database, the data can be examined for quality assurance. These input data sets are later moved into consolidated and geographically referenced master tables of link and TAZ activity data. Finally, these master tables are processed to produce hourly tables and hourly activity data input files for DTIM. 41

43 Table 7 Vintage of travel demand models for link based and traffic analysis zone Metropolitan Planning Organizations AMBAG 2010 Links, Trips 06/15/2015 BCAG 2010 Links, Trips 05/13/2015 FCOG 2008 Links 06/11/2015 CALTRANS 2010 Links, Trips 12/09/2014 KCOG 2010 Links 06/11/2015 KCAG 2010 Links 06/11/2015 MTC 2010 Links, Trips 03/23/2015 MCTC 2010 Links 06/11/2015 MCAG 2010 Links, Trips 06/11/2015 SACOG 2010 Links, Trips 05/08/2014 SANDAG 2008 Links, Trips 12/09/2014 SBCAG 2010 Links, Trips 04/06/2015 SCAG 2008 Links, Trips 01/23/2014 SJCOG 2010 Links, Trips 06/11/2015 SLOCOG 2010 Links, Trips 12/19/2014 StanCOG 2010 Links, Trips 06/11/2015 SCRTPA 2010 Links, Trips 07/13/2015 TCAG 2010 Links 06/11/2015 TMPO 2010 Links, Trips 04/02/2015 Trips data from Caltrans Statewide Travel Demand model were used Spatial Adjustment TDM Version Base year The spatial allocation of county-wide EMFAC emissions is accomplished using gridded, hourly emission estimates from DTIM normalized by county. DTIM uses emission rates from EMFAC along with activity data, digitized roadway segments (links) and traffic analysis zone centroids to calculate gridded, hourly emissions for travel and trip ends. DTIM considers fewer vehicle categories than EMFAC outputs; therefore a mapping between EMFAC and DTIM vehicle categories is necessary. Categories of emissions after running DTIM are presented in Table 8. The categories are represented by the listed source classification codes (SCC) developed by CARB and depend on vehicle type, technology, and whether the vehicle is catalyst, non-catalyst, or diesel. Light- and medium-duty vehicles are separated from heavy-duty vehicles to allow for separate reporting and control strategy applications. Data types received Data received on 42

44 Table 8 DTIM Emission Categories SCC for light and medium duty gas vehicles SCC for heavyduty gas vehicles SCC for light-duty and medium-duty diesel vehicles SCC for heavyduty diesel vehicles Description Catalyst Start Catalyst Running Non-catalyst Start Non-catalyst Hot Soak Diurnal , 508 Diesel Exhaust Running Resting Multi-Day Resting Multi-Day Diurnal , 513, 613, PM Tire Wear , 514, 614, PM Brake Wear Catalyst Buses Non-catalyst , 717 Diesel Bus Catalyst Idle Non-catalyst Idle , 520, 620, Diesel Idle PM Road Dust DTIM and EMFAC2014 are both run using the 13 vehicle types shown in Table 9. In order to obtain better resolved spatiotemporal surrogates, the DTIM runs are split by light-duty (LDA, LDT1, LDT2, MDV, LHDT1, LHDT2, Urban Bus, MH, MCY) and heavy-duty (T6/T7 HHDT, SBUS, Other BUS) vehicle classes, and also by fuel type (gas, diesel). Each DTIM run outputs emissions for categories from 1-13; therefore, the mapping from Error! Reference source not found.9 is used to preserve the spatial surrogates for each of the four DTIM runs. These codes depend on vehicle type, technology and whether the vehicle is catalyst, non-catalyst, or diesel. 43

45 Table 9 Vehicle classification and type of adjustment DTIM Category Vehicle type Type of adjustment 1 LDA LD 2 LDT1 LD 3 LDT2 LD 4 MDV LD 5 LHDT1 LM 6 LHDT2 LM 7 T6 LM 8 T7 HHDT HHDT 9 Other Bus LM 10 School Bus Unadjusted on weekdays, zeroed on weekends 11 Urban Bus LD 12 Motorhomes LD 13 Motorcycles LD Temporal Adjustment (Day-of-Week adjustments to EMFAC daily totals) EMFAC2014 produces average day-of-week (DOW) estimates that represent Tuesday, Wednesday, and Thursday. In order to more accurately represent daily emissions, DOW adjustments are made to enable emissions estimation for other days including Friday, Saturday, Sunday and Monday. The DOW adjustment factors are developed using California Vehicle Activity Database (CalVAD) data. CalVAD, developed by UC Irvine for CARB, is a system that fuses available ground-truth data sources from California Department of Transportation (CalTrans) to produce a best estimate of vehicle activity by vehicle class. The CalVAD data set includes hourly ground-truth measurements of Vehicle Miles Travelled (VMT) on the road network covers all California roadways at a fine spatial (state 4k grid) and temporal (hourly) resolution. However, DOW adjustment filtered out counties that have missing data, malfunctioned detectors, and so on. Therefore, only 34 of the 58 counties had good quality data. In order to fill the missing 24 counties data to cover all of California, a county which is nearby and similar in geography is used as a substitute. The temporal fractions are developed for three categories of vehicles: passenger cars (LD), light- and medium-duty trucks (LM), and heavy-heavy duty trucks (HHDT). Error! Reference source not found.9 also shows the corresponding assignment to 44

46 each vehicle type. Furthermore, the CalVAD fractions are scaled so that a typical workday (Tuesday, Wednesday, or Thursday) gets a scaling factor of 1.0. All other days of the week receive a scaling factor where their VMT is related back to the typical work day. This means there are a total of five weekday scaling factors. Lastly, the CalVAD data were used to create a typical holiday, because the traffic patterns for holidays are quite different than a typical week day. Thus, in the end, there are six daily fractions for each of the three vehicle classes, for all 58 counties. The DOW factors and vehicle type can be found in Appendix A: Day of week redistribution factors by vehicle type and county Temporal Adjustment (Hour-of-Day re-distribution of hourly travel network volumes) The travel networks provided by local transportation agencies and used with DTIM represent an hourly distribution for an average day. As for EMFAC, it is assumed that these average day-of-week hourly distributions represent hourly mid-week activities (i.e. for Tuesday, Wednesday, and Thursday). As such, they lack the temporal variations that are known to occur on other days of the week. To rectify this, the CalVAD data are used to develop hour-of-day profiles for Friday through Monday and a typical holiday. In a similar manner as the DOW factors, these hour-of-day profiles are used to re-allocate the hourly travel network distributions used in DTIM to Friday through Monday and a typical holiday. The hour-of-day profiles can be found in Appendix B: Hour of Day Profiles by vehicle type and county Summary of On-road Emissions Processing Steps Eight steps are used to spatially and temporally allocate EMFAC emissions by hour and grid cell: 1. Activity Data a. EMFAC is run in default mode for a single day to generate hourly activity data for each vehicle type and county: VMT, vehicle population, and 45

47 number of vehicle trips. This is a single day s run, as EMFAC2014 yields the same hourly activity data for every day of the year. b. The activity data are used to generate various input files for ITN and DTIM. The general goal is to determine how much each activity belongs to each vehicle type through the day. 2. Road Network a. Pull a full copy of the California road network from the ITN database, using MPO inputs. b. Convert the ITN results to a form readable by DTIM. c. Apply hourly DOW fractions to travel network volumes. 3. Meteorological Input Data a. Gridded, hourly temperature (T) and relative humidity (RH) are modeled using CALMET. Section 3.1 describes the development of these meteorological (met) data in more detail. b. Daily met files are prepared in formats readable by both EMFAC2014 and DTIM4. 4. EMFAC Emission Rates a. EMFAC is run in emissions rates mode (using monthly-average T and RH) to generate a look-up table of on-road mobile source emission rates by speed, temperature, and relative humidity for each county. These results are created on a monthly-average basis to save processing time. b. The emissions rates are pulled from the EMFAC database and reformatted in the DTIM-ready IRS file format. 5. EMFAC Emissions a. EMFAC is run in emissions mode (using day-specific T and RH) to provide county-wide on-road mobile source emission estimates by day and hour for EMFAC categories. b. These results are saved for later use. 6. DTIM 46

48 a. DTIM is run for one week (five representative days since Tuesday, Wednesday and Thursday are treated as a single day) in the summer and in the winter. b. Convert the DTIM output results into MEDS format for further processing. More details on the DTIM and scaling processing can be found in Appendix C. 7. Scale EMFAC Emissions Using DTIM a. For each day of EMFAC emissions, the closest day-of-week matching DTIM file is chosen for scaling. b. The daily, county-wide EMFAC emissions are distributed spatially and temporally using the DTIM MEDS files as surrogates, as shown by the equation: E P,ij,hr,cat = EF P,daily,cat,cnty DTIM P,ij,hr,cat DTIM P,daily,cat,cnty Where the variables above are defined as: E = grid cell emissions EF = EMFAC emissions DTIM = DTIM emissions p = pollutant i,j = grid cell hr = hourly emissions cat = emission category daily = daily emissions cnty = county c. Finally, the Caltrans day-of-week factors are applied to the gridded, hourly emissions to better match traffic patterns. 8. Final Formatting a. The final step of on-road emissions processing is to convert the gridded, hourly emissions data to a NetCDF (Network Common Data Form) file usable by the CMAQ (Community Multiscale Air Quality) photochemical model. 47

49 Adjustment to the Future Year On-road Emissions ARB is committed to reduce the diesel NOx emissions for medium heavy duty diesel trucks and heavy heavy duty diesel trucks in the San Joaquin Valley for 2024 and The reductions are 18.2 tpd and 18.8 tpd for 2024 and 2025 respectively. The county specific factors are applied to the 2024 and 2025 onroad emissions for medium heavy duty diesel trucks and heavy heavy duty diesel trucks. The factors for 2024 and 2025 are shown in Table 10 and Table 11 respectively. Table 10 County specific factors for 2024 County Factor Fresno Kern King Madera Merced San Joaquin Stanislaus Tulare Table 11 County specific factors for 2025 County Factor Fresno Kern King Madera Merced San Joaquin Stanislaus Tulare

50 3.5 Estimation of Gridded Biogenic Emissions Biogenic emissions were estimated using the Model of Emissions of Gases and Aerosols from Nature (MEGAN) version 2.04 (Guenther, et al., 2006). MEGAN estimates biogenic emissions as a function of normalized emission rates (i.e. emission rates at standard conditions), which are adjusted to reflect variations in temperature, light, leaf area index (LAI), and leaf age (estimated from changes in LAI). The default MEGAN input databases for emission factors (EFs), plant functional types (PFTs), and LAI are not used in the application of MEGAN in California. Instead, California-specific emission factor and PFT databases were translated from those used in the Biogenic Emission Inventory GIS (BEIGIS) system (Scott & Benjamin, 2003) to improve emission estimates and to maintain consistency with previous California biogenic emission inventories. LAI data were derived from the MODIS 8-day LAI satellite product. Hourly surface temperatures were from observations gridded with the CALMET meteorological model and insolation data was provided by the WRF meteorological fields, as discussed in Section 3.1. Emissions of isoprene, monoterpenes, and methylbutenol were estimated from California-specific gridded emission factor data, while emissions of sesquiterpenes, methanol, and other volatile organic compounds were estimated from California-specific PFT data and PFT-derived emission rates. MEGAN emissions estimates for California were evaluated during the California Airborne BVOC Emission Research in Natural Ecosystems Transects (CABERNET) field campaign in 2011 (Karl, et al., 2013), (Misztal, et al., 2014) and were shown to agree to within +/-20% of the measured fluxes (Misztal, et al., 2015), which is well within the stated model uncertainty of 50%. 3.6 Estimation of Ocean-going Vessel (OGV) Emissions As of March 2018, an average-day OGV emission file was provided by an in house CARB OGV model (ARB-PTSD) and ship emissions were allocated corresponding to the vessel traffic lanes. These traffic lanes were estimated from three different sources: National Waterway Network, The Ship Traffic, Energy and Environment Model, and 49

51 Automated instrumentation system (AIS) telemetry data collected in The emission data output from the OGV model contains criteria pollutants as well as fuel consumption. The South Coast Air Quality Management District (SCAQMD) provided port activity data for The weekly port activity for every month of the year were applied to the entire south coast subdomain. After applying the port activity factors mentioned above, emissions were separated by at berth and everything else. At berth emissions are processed through SMOKE and plume rise is calculated for every day of the year Kwok (2015). For transit, maneuvering, and anchorage, emissions are distributed evenly in two vertical layers (2 and 3) as described in Kwok (2015). It is worth nothing that the minimal impact from OGV emissions for the San Joaquin Valley domain is limited to emissions at the port of Stockton and emissions at sea off the coast of San Luis Obispo. 3.7 Estimation of Other Day-Specific Sources Day-specific data were used for preparing base case inventories when data were available. CARB and district staff were able to gather hourly/daily emission information for 1) wildfires and prescribed burns, 2) paved and unpaved road dust, and 3) agricultural burns in six districts (more details highlighted below). Additionally, CARB and district staff reflected residential wood curtailment programs in the base case, reference and future year modeling inventories. In addition, emissions in future years were removed for facilities that have closed after For the reference and future year inventories, which are used to calculate RRFs, dayspecific emissions for wildfires, prescribed burns, and wildland fires use (WFU) are left out of the inventory. All other day-specific data are included in both reference and future year modeling inventories. 50

52 3.7.1 Wildfires and Prescribed Burns Day-specific, base case estimates of emissions from wildfires and prescribed fires were developed in a two part process. The first part consisted of estimating micro-scale, firespecific emissions (i.e. at the fire polygon scale, which can be at a smaller spatial scale than the grid cells used in air quality modeling). The second part consisted of several steps of post-processing fire polygon emission estimates into gridded, hourly emission estimates that were formatted for use in air quality modeling. Fire event-specific emissions were estimated using a combination of geospatial databases and a federal wildland fire emission model, first described in (Clinton, et al., 2006). A series of pre-processing steps were performed using a Geographic Information System (GIS) to develop fuel loading and fuel moisture inputs to the First Order Fire Effects (FOFEM) fire emission model (Lutes, et al., 2012). Polygons from a statewide interagency fire perimeters geodatabase (fire12_1.gdb, downloaded June 4, 2013) maintained by the Fire and Resource Assessment Program (FRAP) of the California Department of Forestry and Fire Protection (CALFIRE) provided georeferenced information on the location, size (area), spatial shape, and timing of wildfires and prescribed burns. Under interagency Memorandums of Understanding, federal, state, and local agencies report California wildfire and prescribed burning activity data to FRAP. Using GIS software, fire polygons were overlaid upon a vegetation fuels raster dataset called the Fuel Characteristic Classification System (FCCS) (Ottmar, et al., 2007). The FCCS maps vegetation fuels at a 30 meter spatial resolution, and is maintained and distributed by LANDFIRE.GOV, a state and federal consortium of wildland fire and natural resource management agencies. With spatial overlay of fire polygons upon the FCCS raster, fuel model codes were retrieved and component areas within each fire footprint tabulated. For each fuel code, loadings (tons/acre) for fuel categories were retrieved from a FOFEM look-up table. Fuel categories included dead woody fuel size classes, overstory live tree crown, understory trees, shrubs, herbaceous vegetation, litter, and duff. Fuel moisture values for each fire were estimated by overlaying fire polygons on year- and month-specific 1 km spatial resolution fuel moisture raster files generated from the national Wildland Fire 51

53 Assessment System (WFAS.net) and retrieving moisture values from fire polygon centroids. Fire event-specific fuel loads and fuel moisture values were compiled and formatted to a batch input file and run through FOFEM. A series of post-processing steps were performed on the FOFEM batch output to include emission estimates (pounds/acre) for three supplemental pollutant species (NH3, TNMHC and N2O) in addition to the seven species native to FOFEM (CO, CO2, PM2.5, PM10, CH4, NOx, SO2), and to calculate total emissions (tons) by pollutant species for each fire. Emission estimates for NH3, TNMHC and N2O were based on mass ratios to emitted CO and CO2 (Gong, et al., 2003). Fire polygon emissions were apportioned to CMAQ model grid cells using area fractions, developed using GIS software, by intersecting fire polygons to the grid domain. Another set of post-processing steps were applied to allocate fire polygon emissions by date and hour of the day. Fire polygon emissions were allocated evenly between fire start and end dates, taken from the fire perimeters geodatabase. Daily emissions were then allocated to hour of day and to the model grid cells and distributed vertically using a method developed by the Western Regional Air Partnership (WRAP), which specifies a pre-defined diurnal temporal profile, plume bottom and plume top for each fire (WRAP, 2005) Paved Road Dust: Statewide emissions from paved road dust were adjusted for each day of the baseline year. The adjustment reduced emissions by 25% from paved road dust on days when precipitation occurred. Paved road dust emissions are calculated using the AP-42 method described in (U.S. EPA, 2006). This methodology includes equations that adjust emissions based on average precipitation in a month; these precipitation-adjusted emissions were placed in the CEIDARS and CEPAM databases. Since daily precipitation totals are readily available, CARB and district staff agreed that paved road dust emissions should be estimated for 52

54 each day rather than by month as described in the AP-42 methodology. The emissions from CEIDARS were replaced with day-specific data. A description of the steps used to calculate day-specific emissions is as follows: Daily uncontrolled emissions for each county/air basin are estimated from the AP-42 methodology [Equation (1) on page ]. No monthly precipitation adjustments are incorporated into the equation to estimate emissions. To adjust for precipitation, daily precipitation data for 2013 were provided by an inhouse database maintained by CARB staff that stores meteorological data collected from outside sources. The specific data sources for these data include: Remote Automated Weather Stations (RAWS), Atmospheric Infrared Sounder (AIRS), California Irrigation Management Information System (CIMIS) networks, SFBMET (San Francisco Bay Area Meteorology), and Federal Aviation Administration (FAA). FAA data provide precipitation data collected from airports in California. If the precipitation is greater than or equal to 0.01 inches (measured anywhere in a county or county/air basin piece on a particular day), then the uncontrolled emissions are reduced by 25% for that day only. This reduction of emissions follows the recommendation in AP-42 as referenced above. Replace the annual average emissions with day-specific emissions for every day in the corresponding emission inventory dataset Unpaved Road Dust: Statewide emissions from unpaved road dust were adjusted for rainfall suppression for each day of the year. The adjustment reduced county-wide emissions by 100% (total suppression) from unpaved road dust on days when precipitation greater than 0.01 occurred in a county/air basin. Dust emissions from unpaved roads were calculated using an emission factor derived from tests conducted by the University of California, Davis, and the Desert Research Institute (DRI). Unpaved road vehicle miles traveled (VMT) were based on county-specific road mileage estimates. 53

55 Emissions were assumed to be suppressed for each day with rainfall of 0.01 inch or greater using equation (2) from the method described in (U.S. EPA, 2006). The equation adjusts emissions based on annual precipitation; these precipitation-adjusted emissions were placed in the CEIDARS database. Similar to paved road dust, CARB and district staff agreed that unpaved road dust emissions should be estimated for each day. The emissions from CEIDARS were replaced with day-specific data for the appropriate years. Following is a description of the steps that were taken to calculate day-specific emissions. Start with the daily uncontrolled emissions for each county/air basin as estimated from CARB s methodology. In other words, no precipitation adjustments have been incorporated in the emission estimates. Use the same daily precipitation data as for paved road dust (see above) If the precipitation is greater than or equal to 0.01 inches measured anywhere in a county or county/air basin portion on a particular day, then the emissions are removed for that day only. Replace the annual average emissions with day-specific emissions for every day Agricultural Burning: Agricultural burning day-specific emission estimations were incorporated into the inventory for the following areas: San Joaquin Valley The San Joaquin Valley Air Pollution Control District estimated emissions for each day of 2013 when agricultural burning occurred. Emissions were estimated for the burning of pruning, field crops, weed abatement, and other solid fuels. Information needed to estimate emissions came from the district s Smoke Management System, which stores information on burn permits issued by the district. In order to obtain a daily burn authorization, the person requesting the burn provides information to the district, including the acres and type of material to be burned, the specific location of the burn, 54

56 and the date of the burn. Acres are converted to tons of fuel burned using a fuel loading factor based on the specific crop to be burned. Emissions are calculated by multiplying the tons of fuel burned by a crop-specific emission factor. More information can be found in (ARB-Miscellaneous Methodologies, 2013). To determine the location of the burn, district staff created spatial allocation factors for each 4 kilometer grid cell used in modeling. These factors were developed for burn zones in the San Joaquin Valley based on the agricultural land coverage. Daily emissions in each agricultural burn zone were then distributed across the zone/grid cell combinations using the spatial allocation factors. Emissions were summarized by grid cell and day. Burning was assumed to occur over three hours from 10:00 a.m. to 1:00 p.m., except for two categories. Orchard removals were assumed to burn over eight hours from 10:00 a.m. to 6:00 p.m. Vineyard removals were assumed to burn over five hours from 10:00 a.m. to 3:00 p.m. Sacramento Sacramento Metropolitan Air Quality Management District provided information needed to calculate emissions in Sacramento County from agricultural burning for each day of 2013 when agricultural burning occurred. Using the same methodology as San Joaquin Valley, emissions were estimated for the burning of prunings, field crops, weed abatement and other solid fuels. Information needed to estimate emissions came from burn permits issued by the district. In order to obtain a burn permit, the person requesting the burn provides information to the district, including the acres to be burned, the specific location of the burn and the date of the burn. Acres are converted to tons of fuel burned using a fuel loading factor based on the specific crop to be burned. Emissions are calculated by multiplying the tons of fuel burned by a crop-specific emission factor. The location of the burn was converted to latitude/longitude based on the address or description of location provided by the burn permit holder, then ultimately to grid cell. Burning was assumed to occur over eight hours from 10:00 a.m. to 6:00 p.m. 55

57 Yolo-Solano Yolo-Solano Air Quality Management District provided information needed to calculate emissions from agricultural burning for each day of 2013 when agricultural burning occurred. Data were provided for their region: all of Yolo County and the Sacramento Valley portion of Solano County. Using the same methodology as San Joaquin Valley, emissions were estimated for the burning of prunings, field crops, weed abatement and range improvement. The location of the burn was converted to latitude/longitude based on the address or description of location provided by the burn permit holder, then ultimately to grid cell. Burning was assumed to occur over five hours from 11:00 a.m. to 4:00 p.m. Feather River Feather River Air Quality Management District provided information needed to calculate emissions from agricultural and prescribed burning for each day of 2013 when agricultural burning occurred. Data were provided for Sutter and Yuba Counties. Using the same methodology as San Joaquin Valley, emissions were estimated for the burning of prunings, field crops, weed abatement and other solid waste. The location of each burn was converted to latitude/longitude based on the address or description of location provided by the burn permit holder, then ultimately to grid cell. Orchard prunings were assumed to occur from 9:00 a.m. to 4:00 p.m. The burning of field crops, rice, weeds and ditch banks were assumed to occur from 10:00 a.m. to 5:00 p.m. from March 1 through August 31 and from 10:00 a.m. to 4:00 p.m. from September 1 through February 29. Prescribed burns over 10 acres were assumed to occur from 9:00 a.m. to 12:00 a.m. while prescribed burns less than 10 acres were assumed to occur from 9:00 a.m. to 6:00 p.m. Ventura Ventura County Air Pollution Control District provided emissions in Ventura County from agricultural burning for each day of 2013 when agricultural burning occurred. Using the same methodology as San Joaquin Valley, emissions were estimated for the burning of prunings, field crops, weed abatement, range improvement and prescribed burns not 56

58 included in the wildfires / prescribed burns discussed in the San Joaquin Valley portion of Section Information needed to estimate emissions came from burn permits issued by the district. In order to obtain a burn permit, the person requesting the burn provides information to the district, including the acres to be burned, the specific location of the burn and the date of the burn. Acres are converted to tons of fuel burned using a fuel loading factor based on the specific crop to be burned. Emissions are calculated by multiplying the tons of fuel burned by a crop-specific emission factor. The location of the burn was converted to latitude/longitude based on the address or description of location provided by the burn permit holder, then ultimately to grid cell. Burning was assumed to occur over three hours from 9:00 a.m. to 12:00 p.m Residential Wood Curtailment Emissions were reduced to reflect residential wood curtailment days (no burn days) in three districts: San Joaquin Valley APCD, South Coast AQMD, and Sacramento Metropolitan AQMD. As of March 2018, there are two major changes in the SMOKE processing of RWC, and one major change in future-year curtailment. The first change in the SMOKE-processing of RWC is in temporal allocation. In the past, SMOKE temporally allocated RWC emissions with monthly, weekly and diurnal profiles provided by CARB planning staff. Now the profiles are replaced with the ones based on modeled ambient temperature from WRF with respect to the reference model year (2013). Specifically, a pre-smoke utility program called GentPro is used to generate county-specific temporal profiles taking into account average temperature by grid cell (Manual 3.1, Manual 4.0, Kwok 2016a). Emissions for any given county will only be allocated whenever the daily average temperature by grid cell is below 50 F. In addition, the diurnal profile has also changed. In previous versions the profile consisted of morning-evening peaks however now the profile reflects evening-only activities beginning from 7pm and ending at midnight, with each hour carrying an equal weight. The second change in the SMOKE-processing of RWC is in spatial allocation. A new spatial surrogate for fireplaces was constructed based on the population of houses, 57

59 apartments, and any other residential dwellings with fireplaces (see section 2.3.7). This surrogate is applied to both woodstoves and fireplaces emissions in SMOKE. The change in RWC curtailment programs is only for San Joaquin Valley (SJV); the corresponding programs remain unchanged for South Coast (SC) and Sacramento Valley (SACV). The following describes the current curtailment programs for SC, SACV, and SJV as well as proposed changes to SJV s curtailment program. 58

60 San Joaquin Valley In San Joaquin Valley (SJV), current RWC curtailment programs for base year 2013 and future years 2020, 2024 and 2025 are in effect. The programs are also referred to as Rule Additional RWC reductions are expected as areas of gas utility accessibility increase and a woodstove swap-out program (Burn Cleaner program) rolls out. Here, we summarize the current curtailment programs before describing the new woodstove swap-out program in detail. Current program Base Year (2013): SJVAPCD staff provided the dates in 2013 when a residential wood curtailment was declared based on the October 2008 district rule When observed PM2.5 reached or exceeded 35 µg/m 3, the curtailment was declared. Consequently, emissions were reduced by 65% (i.e. 35% remaining) in the appropriate geographic regions (see Section 2.3.7). Future Years (2020, 2024 and 2025): RWC in future years reflect the latest revision to Rule 4901, based on a September 2014 three-level curtailment program: Level 0 burning allowed Level 1 burning permitted by cleaner-burning woodstoves only Level 2 no burning The consecutive levels are partitioned by values called cut-points, which are also based on the observed PM2.5 concentrations. For example, cut-points denotes the observed PM2.5 at 20 µg/m 3 and 65 µg/m 3, respectively. Cut-point-20 applies the Level 1 curtailment, whereas cut-point-65 applies a more restrictive Level 2 curtailment. 59

61 Updates to the Current Program The SJV RWC curtailment program has been updated to include the Burn Cleaner program. The Burn Cleaner program is applied to the uncurtailed SJV RWC emissions inventory prior to application of Rule In the Burn Cleaner program, the SJVAPCD staff identified hotspots within the SJV air basin as shown in Figure 7. The hotspots are either new areas of gas utility or areas deemed to have persistently poor air quality. The SJVAPCD provides Burn Cleaner reduction factors (or equivalently retention factors) for both the hotspots and the remaining areas, as shown in Table 12. These factors are applied to registered woodstoves only; fireplace emissions are subject to 97% compliance at Level 1 or above. Figure 7 Hot spot areas in San Joaquin Air Basin 60

62 Table 12 County specific burn cleaner retention factors County Number County Name Hotspot Retention Non-hotspot Retention 10 Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Based on the remaining emissions after the Burn Cleaner reductions are applied, county-specific curtailment is then determined. However, the county-specific curtailment criteria are defined differently for hotspots versus non-hotspots areas. Hotspot areas have curtailment on days where observed PM2.5 is greater than or equal to 12 µg/m 3, prompting the Level 1 measure. If PM2.5 hits 35 µg/m 3, then the Level 2 measure is triggered. For non-hotspot areas, curtailment occurs on days where PM2.5 is greater than 20 µg/m 3, prompting a level 1 measure, or greater than 65 µg/m 3 prompting a level 2 measure. For a level 2 curtailment, declared measured emissions were reduced by 97% (i.e. 3% remaining) in the appropriate geographic regions. Greater reductions due to curtailment are assumed in the future years to reflect increased public awareness and thus greater compliance with district rules. To avoid double-counting emission reductions on curtailment days, the modeling inventories were only grown without the control profile applied. Since emissions from RWC have flat growth, the same reductions are used for all future years (2015 and later). 61

63 South Coast SCAQMD staff provided the dates in 2013 when a residential wood combustion curtailment (RWCC) was declared based on district rule 455. When a RWCC was declared emissions were reduced by 75% (i.e. 25% remaining) in the appropriate geographic regions (see Section 2.3.7). In future years, emissions continued to be reduced by 75%, using the same dates as in Sacramento SMAQMD staff provided the dates in 2013 when a RWCC was declared based on district rule 421. Per this rule, a mandatory curtailment (no burning) is called when: Stage 1: the 24-hour average PM2.5 concentration may exceed 31 μg/m 3 but is not likely to exceed 35 μg/m 3 Stage 2: the 24-hour average PM2.5 concentration may exceed 35 μg/m 3 When a RWC was declared, emissions in Sacramento County (see Section 2.3.7) were reduced as follows: For Stage 1: 57% (i.e. 43% remaining) For Stage 2: 70% (i.e. 30% remaining) In future years, emissions were continued to be reduced by 57% and 70% for stage 1 and stage 2, respectively. The same calendar dates from 2013 were used in future years. 62

64 3.7.6 Estimation of Agricultural Ammonia Emissions: Ammonia emissions from fertilizers/pesticides (EIC3 530) and livestock (EIC3 620) are separated from the aggregated area source inventory as they are affected by local meteorology. In previous work, a flat temporal profile was assigned to both sectors but due to the dependence on meteorology factors a more realistic approach to temporal representation was needed. For EIC3 530, the depending factors are WRF s two-meter temperature and ten-meter wind speed. For EIC3 620, the factors are WRF s ground temperature and aerodynamic resistance. Through GenTpro these meteorological factors are averaged by county before creating year-long hourly profiles for each of the respective sectors. All algorithms are described in the SMOKE Manual 4.0, while the results of CARB in-house tests are presented by Kwok (2016) Closed Facilities Emissions in future years were removed for facilities that have closed beyond the reference year. In other words, the emissions were removed from future year inventories for a facility that was included in the 2012 inventory but stopped operating after Local air district staff members provided the lists of these facilities. 3.8 Application of Control Measure Reduction Factors in San Joaquin Valley Controls were applied to reduce emissions in the future year attainment modeling inventories for 2020, 2024, and Control strategies for targeted emissions categories (e.g. charbroiling) are outlined in Sections to A summary of the control strategies applied to each future year is described in Sections and Charbroiling Control strategies to reduce PM2.5 emissions from under-fired charbroilers are expected to be achieved by: Funding 100% of equipment and installation cost plus two years operation and maintenance cost of emissions control systems for 30% (Vanderspek, 2017) of 63

65 existing restaurants with charbroilers within urban boundaries in hot spot areas. (SJVUAPCD, 2017) Funding 50% of equipment and installation cost plus two years (Vanderspek, 2017) operation and maintenance cost of emissions control systems for large new restaurants with charbroilers within urban boundaries of hot spot areas (SJVUAPCD, 2017). PM2.5 reductions are applied to Hot Spot regions as identified in Error! Reference source not found.. Figure 8 Hot Spot Areas for application of under-fired charbroiling PM2.5 reductions. 64

66 3.8.2 Residential Wood Combustion Control strategies to reduce emissions from residential wood combustion (RWC) were applied in accordance with future year curtailment rules described in Section Error! Reference source not found Future Year 2020 The only controls applied on top of the projected 2020 inventory emissions were to RWC. Prior to reduction, the RWC emissions were allocated to locations by the new fireplace surrogates. Subsequently, the RWC emissions were reduced according to future year baseline curtailment rules using a compliance rate of 97% (refer to Section Error! Reference source not found.) Future Years 2024 and 2025 Residential Wood Combustion For future year attainment modeling of 2024 and 2025, RWC emissions are subject to more stringent controls. RWC emissions are reduced per the Burn Cleaner program as explained in Section Error! Reference source not found.. The value of the RWC emissions reductions is expected to the same for 2024 and 2025 given the lack of growth in RWC emissions, and the application of the curtailment is the same in both years. Charbroiling For both future years 2024 and 2025, emissions from under-fired charbroilers are reduced in Hot Spot regions (described in Section 3.8.1). Other reductions In addition to reductions from RWC and charbroiling, control strategies are applied to other sources in order to further reduce NOx and PM2.5 emissions in future years 2024 and 2025 (refer to Table Table 13 or a list of these sources). 65

67 66

68 Table 13 NOx and PM2.5 other sources reduced Reduction Source NOx PM2.5 Electrification of ag IC engines (50% reduction) X X Reduce stationary source fuel combustion X Reduce ag equipment X X Reduce off-road equipment X Reduce locomotives X Reduce heavy duty diesel trucks X Reduce flaring operations (20% reduction) X Enhanced conservation management practices (tillage) X Enhanced conservation management practices (fallow land) X Reduce urban dust in Bakersfield (25% reduction) X Reduce woodchips at Bakersfield-Planz monitor (75% reduction) X 67

69 4 Quality Assurance of Modeling Inventories As mentioned in Section 1.3, base case modeling is intended to demonstrate confidence in the modeling system. Quality assurance of the data is fundamental in order to detect any possible outliers and potential problems with emission estimates. The most important quality assurance checks of the modeling emissions inventory are summarized in the following sections. 4.1 Area and Point Sources Before utilizing SMOKE to process the annual emissions totals into temporally, chemically, and spatially-resolved emissions inventories for photochemical modeling, all SMOKE inputs are subject to extensive quality assurance procedures performed by CARB staff. Annual and forecasted emissions are carefully reviewed before input into SMOKE. CARB and district staff review data used to calculate emissions along with other associated data, such as the location of facilities and assignment of SCC to each process. Growth and control information are reviewed and updated as needed. The next check is to compare annual average emissions from CEPAM with planning inventory totals to ensure data integrity. The planning and modeling inventories start with the same annual average emissions. The planning inventory is developed for an average summer day and an average winter day, whereas the modeling inventory is developed by month. Both inventory types use the same temporal data described in Section 2.2. The summer planning inventory uses the monthly throughputs from May through October. Similarly, the winter planning inventory uses the monthly throughputs from November through April. The modeling inventory produces emissions for a weekday, Saturday and Sunday for each month. Annual emissions totals are plotted using the same gridding inputs as used in SMOKE in order to visually inspect and analyze the spatial allocation of emissions independent of temporal allocation and chemical speciation. Spatial plots by source category like the one shown in Figure 9 Example of a spatial plot by source categoryfigure 9 are carefully screened for proper spatial distribution of emissions. 68

70 Figure 9 Example of a spatial plot by source category Before air quality model-ready emissions files are generated by SMOKE, the run configurations and parameters set within the SMOKE environment are checked for consistency for both the reference and future years. To aid in the quality assurance process, SMOKE is configured to generate inventory reports of temporally, chemically, and spatially-resolved emissions inventories. CARB staff utilize the SMOKE reports by checking emissions totals by source category and 69

71 region, creating and analyzing time series plots, and comparing aggregate emissions totals with the pre-smoke emissions totals obtained from CEPAM. A screenshot capture of a portion of such report can be seen in Figure 10. Figure 10 Screen capture of a SMOKE-generated QA report Area and Point Sources Temporal Profiles Checks for missing or invalid temporal assignments are conducted to ensure accurate temporal allocation of emissions. Special attention is paid to checking monthly throughputs and appropriate monthly temporal distribution of emissions for each source category. In addition, checks for time-invariant temporal assignments are done for certain source categories and suitable alternate temporal assignments are determined and applied. For the agricultural source sector (e.g. agricultural pesticides/fertilizers, farming operations, fugitive windblown dust, managed burning and disposal, and farm equipment), replacement temporal assignments are extracted from the Agricultural Emissions Temporal and Spatial Allocation Tool (AgTool) (Anderson, et al., 2012). The AgTool is a database management system capable of temporally and spatially allocating emissions from the agricultural source sector. It was developed by Sierra Research, Inc. and its subcontractor Alpine Geophysics, LLC along with collaboration from CARB and the San Joaquin Valley Air Pollution Control District (SJVAPCD). Temporal allocation data outputs from the AgTool, were compiled 70

72 using input data provided by the UC Cooperative Extension, U.S. Department of Agriculture (USDA), and the CA Department of Pesticide Regulation (DPR). Further improvements to temporal profiles used in the allocation of area source emissions are performed using suitable alternate temporal assignments determined by CARB staff. Select sources from manufacturing and industrial, degreasing, petroleum marketing, mineral processes, consumer products, residential fuel combustion, farming operations, aircraft, and commercial harbor craft sectors are among the source categories included in the application of adjustments to temporal allocation. 4.2 On-road Emissions There are several processes to conduct quality assurance of the on-road mobile source modeling inventory at various stages of the inventory processing. The specific steps taken are described below: 1. Generate an ITN spatial plot to check if there were any missing network activities. 2. Generate a time series plot for each county to check the diurnal pattern of network activities. 3. Generate time series plots for the DTIM output files by county and by SCC to check the diurnal pattern. 4. Generate time series plots for the on-road mobile source files after scaling to EMFAC 2014 emissions (MEDS files) by county and SCC to check the diurnal pattern. 5. Compare the statewide daily total emissions for the MEDS files and the EMFAC 2014 emissions files to ensure that the emissions are the same. 6. Generate the spatial plot for the MEDS file to check if there were any missing emissions. 7. Generate time series and spatial plots again to check the final MEDS files. 71

73 4.3 Day-specific Sources Wildfires and Prescribed Burns GIS records for 364 wildfires and 125 prescribed wildland burn events reported for 2013 were downloaded from and imported to a geodatabase. Data fields included wildfire or burn project name, burned area, start and end dates. A series of geoprocessing steps were used to map and overlay wildfire and prescribed burn footprint polygons on the statewide vegetation fuels (FCCS) and moisture raster datasets, to retrieve associated fuel loadings and moisture values for use as input to FOFEM. Wildfire and prescribed burn footprint polygons were also overlaid on the statewide 4-km modeling grid to assign grid cell IDs to each wildfire and prescribed burn. Emission estimates for each wildfire and prescribed burn event were generated by FOFEM and summarized in an Access database Paved Road Dust The average daily emissions inventory was adjusted with day-specific precipitation data to produce a day-specific emissions inventory. Total emissions by county before the adjustment were compared to CEPAM for a reasonable match. After the adjustment, the day-specific total emissions by county were compared to CEPAM using time series plots. These plots were verified to confirm that there were only two values for every county/air basin/district: high values and low values. The high values are emissions that were not affected by rain adjustment, while the low values are emissions that were affected by the 25% rain adjustment reduction. Additionally the day-specific total was also compared to other inventory years to verify the expected growth trend Unpaved Road Dust Unpaved road dust followed the same quality assurance process as paved road dust. The reduction efficiency for unpaved roads is increased to 100% on precipitation days. 72

74 4.3.4 Agricultural Burning Checks were done to verify the quality of the agricultural burn data. The day-specific emissions from agricultural burning were compared to the emissions from CEPAM for each county to check for reasonableness. Time series plots were reviewed for each county to see that days when burning occurred matched the days provided by the local air district. For each county, a few individual fires were calculated by hand starting from the raw data through all the steps to the final MEDS files to make sure the calculations were done correctly. Spatial plots were made to double check the locations of each burn. 4.4 Additional QA In addition to the QA described above, comparisons are made between annual average inventories from CEPAM and modeling inventories. The modeling inventory shows emissions by month and subsequently calculates the annual average for comparison with CEPAM emissions. Annual average inventories and modeling inventories can be different, but differences should be well understood. For example, modeling inventories are adjusted to reflect different days of the week for on-road motor vehicles as detailed in Section 3.4; since weekend travel is generally less than weekday travel, modeling inventory emissions are usually lower when compared to annual average inventories from CEPAM. Figure 11 provides a screen capture of a report that summarizes different emission categories for San Luis Obispo County. Please note that this table is only an example since emissions have been updated from what is displayed here. 73

75 County:40 Spec:NOx EIC Description Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Annual CEPAM Difference 10 electric utilities cogeneration oil and gas production (combustion) petroleum refining (combustion) manufacturing and industrial food and agricultural processing service and commercial other (fuel combustion) sewage treatment landfills incinerators soil remediation other (waste disposal) laundering degreasing coatings and related process solvents printing adhesives and sealants other (cleaning and surface coatings) oil and gas production petroleum refining petroleum marketing other (petroleum production and marketing) chemical food and agriculture mineral processes metal processes wood and paper glass and related products electronics other (industrial processes) consumer products architectural coatings and related process sol pesticides/fertilizers asphalt paving / roofing residential fuel combustion farming operations construction and demolition paved road dust unpaved road dust fugitive windblown dust fires managed burning and disposal cooking other (miscellaneous processes) on-road vehicles aircraft trains ships and commercial boats ocean going vessels commercial harbor craft recreational boats off-road recreational vehicles off-road equipment farm equipment fuel storage and handling geogenic sources *** Total Notes: CEPAM refers to annual average emissions from 2016 SIP Baseline Emission Inventory Tool with external adjustments: Monthly gridded emissions comes from GeoVAST mo-yr/avg tabular summary - gid 319 On-road vehicles: The modeling inventory adjusts on-road by day of week as well as day-specific temperatures and relative humidity - Fridays are higher wit time series plots shows weekdays are ~9-10 tpd Trains: The modeling inventory reflects the revised locomotive emissions; the planning inventory reflects the previous emission estimates OGV model produces gridded OGV emissions, which can vary from planning inventory (these emissions include OC1 and OC2 offshore air basins) CHC The modeling inventory reflects the revised commercial harbor craft emissions; the planning inventory reflects the previous emission estimates Figure 11 Screenshot of comparison of inventories report 74

76 Staff also review how modeling emissions vary over a year. Figure 12 provides an example of a modeling inventory time series plot for San Luis Obispo County for areawide sources, on-road sources and off-road sources. Again, this figure is only an example. Figure 12 Daily variation of NOx emissions for mobile sources for San Luis Obispo 75

77 4.5 Model ready files QA Prior to developing the modeling inventory emissions files used in the photochemical models, the same model-ready emissions files developed for the individual source categories (e.g. on-road, area, point, day-specific sources) are checked for quality assurance. Extensive quality assurance procedures are already performed by CARB staff on the intermediate emissions files (e.g. MEDS, SMOKE-generated reports), however, further checks are needed to ensure data integrity is preserved when the model-ready emissions files are generated from those intermediate emissions files. Comparisons of the totals for both the intermediate and model-ready emissions files are made. Emissions totals are aggregated spatially, temporally, and chemically to singlelayer, statewide, daily values by inventory pollutant. Spatial plots are also generated for both the intermediate and model-ready emissions files using the same graphical utilities and aggregated to the same spatial, temporal, and chemical resolution to allow equal comparison of emissions. Any discrepancies in the emissions totals are reconciled before proceeding with the development of the model-ready inventory emissions files. Before combining the model-ready emissions files of the individual source category inventories into a single model-ready inventory, they are checked for completeness. Day-specific source inventories (when necessary) should have emissions for every day in the modeling period. Likewise, source inventories with emissions files that use averaged temporal allocation (e.g. day-of-week, weekday/weekend, monthly) should have model-ready emissions files to represent every day in the modeling period. In particular, it is important that during these checks source inventories with missing files are identified and resolved. Once all constituent source inventories are complete, they are used to develop the model-ready inventory used in photochemical modeling. When the modeling inventory files are generated, log files are also generated documenting the constituents of each daily model-ready emissions file as an additional means of verifying that each daily model-ready inventory is complete. 76

78 Bibliography Anderson, C., Carlson, T., Dulla, R. & Wilkinson, J., AGTOOL Agricultural sector emissions allocation model Data development. CCOS_11-2, Sacramento, CA: Sierra Research Inc. ARB, Heavy Duty Diesel Vehicle Exhaust PM Speciation Profile. [Online] Available at: df [Accessed August]. ARB, SIPIWG2. [Online] Available at: ARB, Modeling Emissions Inventory Protocol _Imperial, s.l.: s.n. ARB-Miscellaneous Methodologies, District Miscellaneous Processes Methodologies - Managed Burning and Disposal. [Online] Available at: [Accessed ]. ARB-MSEI, Mobile Source Emissions Inventory -- Categories. [Online] Available at: [Accessed ]. ARB, n.d. SIPIWG2. [Online] Available at: ARB-PTSD, Appendix D: Emissions estimation methodology for Ocean-going vessels. [Online] Available at: [Accessed ]. Baek, B. a. S. C., Final Summary Report: Development of SMOKE version 4.0. Under contract ITS:12-764, Sacramento, CA: s.n. 77

79 Caldwell, P. C. H. -N. B. D. a. B. G., Evaluation of a WRF dynamical downscaling simulation over California. Climatic change, 95(3), pp Clinton, N., Gong, P. & Scott, K. 2., Quantification of pollutants emitted from very large wildland fires in Southern California. Atmospheric Environment, Volume 40, pp Dun and Bradstreet, Market insight database.. [Online] Available at: [Accessed 15 June 2015]. FactFinder, U. C. B. /. A., "B25040: House heating fue ACS 5-year estimatesl". [Online] Available at: [Accessed March 2015]. Fovell, R., Validation of WRF Surface meteorology Simulations, presentation November 4th 2008., Sacramento: CCOS Technical Commitee meeting, California Air Resources Board. Funk, t., Stiefer, P. & Chinkin, L., Development of gridded spatial allocation factors for the state of California., Sacramento: STI. Funk, t., Stiefer, P. & Chinkin, L., Development of gridded spatial allocation factors for the state of California., Sacramento: STI. Gong, P., Clinton, N. & Pu, R. T. Y. a. S. J., Extension and input refinement to the ARB wildland fire emissions estimation model. Final report, contract number , Sacramento, CA.: Air Resources Board. Guenther, A. et al., Estimates of global terrestrial isoprene emissions using MEGAN (Model of Emissions of Gases and Aerosols from Nature. Chem. Phys., 6(11), pp

80 Homer, C. et al., Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogrammetric Engineering and Remote Sensing, 81(5), pp Jackson, B., Preparation of a gridded emissions Inventory for Mexico. [Online] Available at: [Accessed ]. Karl, T. et al., Airborne flux measurements of BVOCs above Californian oak forests: experimental investigation of surface and entrainment fluxes, OH densities and Dahmkohler numbers. J. Atmos. Sci., 70(10), pp Kwong, J., Update of the gridded emissions inventory for northern Mexico, s.l.: ARB. Lutes, D., Keane, R. & Reinhardt, E. 2., FOFEM 6 User Guide. USDA Forest Service, Rocky Mountain Research Station, Missoula Fire Sciences Laboratory, Missoula, MT 59808: s.n. Mc. Rae, G. J., A simple procedure for calculating atmospheric water vapor concentration. J. Air Pollution Control Association, Volume 30, p Misztal, P. et al., Evaluation of regional isoprene emission estimates in California based on direct airborne flux measurements. Submitted to Atmos. Chem. Phys.. Misztal, P. K. et al., Airborne flux measurements of biogenic volatile organic compounds over California. Atmos. Chem. Phys., Volume 14, pp Ottmar, R., Sandberg, D. & Riccardi, C. a. P. S., An overview of the Fuel Characteristic Classification System Quantifying, classifying, and creating fuelbeds for resource planning.. Canadian Journal of Forest Research, 37(12), pp Reid, S., Penfold, B. & Chinkin, L., Emission inventory for the Central California Ozone Study (CCOS) review of spatial variations of area, non-road mobile, and point sources of emissions, Petaluma, CA: STI. 79

81 Scott, K. & Benjamin, M., Development of a biogenic volatile organic compounds emission inventory for the SCOS97-NARSTO domain. Atmos. Environ., Volume 37, Supplement 2, pp SJVUAPCD, Intem Number 8: Provide Direction on Pursuing Targeted Strategies Focused on Regional Hot-Spots to Expedite Attainment of Federal PM2.5 Standards in a More Cost Effective Fashion. [Online] Available at: final/08.pdf Systems Applications Inc., DTIM 4 User's Guide. Prepared for the California Department of Transportation, Sacramento: Systems Applications, Inc.. U.S. EPA, AP 42, Compilation of Air Pollutant Emission Factors. In: Volume 1. Chapter 13: Miscellaneous Sources. Section (Paved Road). Fifth ed. s.l.:u.s. EPA. U.S. EPA, Draft Modeling Guidance for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and RegionalHaze, North Carollina: U.S. EPA. U.S. EPA, [Online] Available at: [Accessed 16 August 2016]. Valade, A., Evaluation of the simulation of the 2006 California heat wave by WRF using satellite, Davis, California: U.C Davis. Vanderspek, S. (. C., [Online]. WRAP, Fire Emission Inventory for the WRAP Region Phase I Essential Documentation. [Online] Available at: 80

82 _ pdf [Accessed ]. 81

83 Appendix A: Day of week redistribution factors by vehicle type and county The factors shown in 13 represent the day of week factors for each county for a broad vehicle class: LD is Light Duty, LM is Light and Medium Duty Trucks, and HH is Heavy- Heavy Duty Trucks. Table 14 Day of week adjustment by vehicle class and county County Day of Week LD LM HH Alameda Sunday Alameda Monday Alameda Tues/Wed/Thurs Alameda Friday Alameda Saturday Alameda Holiday Alpine Sunday Alpine Monday Alpine Tues/Wed/Thurs Alpine Friday Alpine Saturday Alpine Holiday Amador Sunday Amador Monday Amador Tues/Wed/Thurs Amador Friday Amador Saturday Amador Holiday Calaveras Sunday Calaveras Monday Calaveras Tues/Wed/Thurs Calaveras Friday Calaveras Saturday Calaveras Holiday Colusa Sunday Colusa Monday Colusa Tues/Wed/Thurs Colusa Friday Colusa Saturday Colusa Holiday Contra Costa Sunday Contra Costa Monday Contra Costa Tues/Wed/Thurs Contra Costa Friday Contra Costa Saturday Contra Costa Holiday El Dorado Sunday El Dorado Monday El Dorado Tues/Wed/Thurs El Dorado Friday El Dorado Saturday El Dorado Holiday Fresno Sunday Fresno Monday Fresno Tues/Wed/Thurs Fresno Friday Fresno Saturday Fresno Holiday Glenn Sunday Glenn Monday Glenn Tues/Wed/Thurs Glenn Friday Glenn Saturday Glenn Holiday Inyo Sunday Inyo Monday Inyo Tues/Wed/Thurs Inyo Friday Inyo Saturday Inyo Holiday Kern Sunday Kern Monday

84 County Day of Week LD LM HH Kern Tues/Wed/Thurs Kern Friday Kern Saturday Kern Holiday Kings Sunday Kings Monday Kings Tues/Wed/Thurs Kings Friday Kings Saturday Kings Holiday Los Angeles Sunday Los Angeles Monday Los Angeles Tues/Wed/Thurs Los Angeles Friday Los Angeles Saturday Los Angeles Holiday Madera Sunday Madera Monday Madera Tues/Wed/Thurs Madera Friday Madera Saturday Madera Holiday Mariposa Sunday Mariposa Monday Mariposa Tues/Wed/Thurs Mariposa Friday Mariposa Saturday Mariposa Holiday Merced Sunday Merced Monday Merced Tues/Wed/Thurs Merced Friday Merced Saturday Merced Holiday Mono Sunday Mono Monday Mono Tues/Wed/Thurs Mono Friday Mono Saturday Mono Holiday Monterey Sunday Monterey Monday Monterey Tues/Wed/Thurs Monterey Friday Monterey Saturday Monterey Holiday Napa Sunday Napa Monday Napa Tues/Wed/Thurs Napa Friday Napa Saturday Napa Holiday Placer Sunday Placer Monday Placer Tues/Wed/Thurs Placer Friday Placer Saturday Placer Holiday Sacramento Sunday Sacramento Monday Sacramento Tues/Wed/Thurs Sacramento Friday Sacramento Saturday Sacramento Holiday San Benito Sunday San Benito Monday San Benito Tues/Wed/Thurs San Benito Friday San Benito Saturday San Benito Holiday San Joaquin Sunday San Joaquin Monday San Joaquin Tues/Wed/Thurs San Joaquin Friday

85 County Day of Week LD LM HH San Joaquin Saturday San Joaquin Holiday San Luis Obispo Sunday San Luis Obispo Monday San Luis Obispo Tues/Wed/Thurs San Luis Obispo Friday San Luis Obispo Saturday San Luis Obispo Holiday Santa Barbara Sunday Santa Barbara Monday Santa Barbara Tues/Wed/Thurs Santa Barbara Friday Santa Barbara Saturday Santa Barbara Holiday Santa Clara Sunday Santa Clara Monday Santa Clara Tues/Wed/Thurs Santa Clara Friday Santa Clara Saturday Santa Clara Holiday Santa Cruz Sunday Santa Cruz Monday Santa Cruz Tues/Wed/Thurs Santa Cruz Friday Santa Cruz Saturday Santa Cruz Holiday Solano Sunday Solano Monday Solano Tues/Wed/Thurs Solano Friday Solano Saturday Solano Holiday Sonoma Sunday Sonoma Monday Sonoma Tues/Wed/Thurs Sonoma Friday Sonoma Saturday Sonoma Holiday Stanislaus Sunday Stanislaus Monday Stanislaus Tues/Wed/Thurs Stanislaus Friday Stanislaus Saturday Stanislaus Holiday Sutter Sunday Sutter Monday Sutter Tues/Wed/Thurs Sutter Friday Sutter Saturday Sutter Holiday Tehama Sunday Tehama Monday Tehama Tues/Wed/Thurs Tehama Friday Tehama Saturday Tehama Holiday Tulare Sunday Tulare Monday Tulare Tues/Wed/Thurs Tulare Friday Tulare Saturday Tulare Holiday Tuolumne Sunday Tuolumne Monday Tuolumne Tues/Wed/Thurs Tuolumne Friday Tuolumne Saturday Tuolumne Holiday Ventura Sunday Ventura Monday Ventura Tues/Wed/Thurs Ventura Friday Ventura Saturday Ventura Holiday

86 County Day of Week LD LM HH Yolo Sunday Yolo Monday Yolo Tues/Wed/Thurs Yolo Friday Yolo Saturday Yolo Holiday

87 Appendix B: Hour of Day Profiles by vehicle type and county The factors shown in Table 15 represent the day of week factors for each county for a broad vehicle class: LD is Light Duty, LM is Light and Medium Duty Trucks, and HH is Heavy- Heavy Duty Trucks. Table 15 Hour of Day Profiles by vehicle type and county Alameda Alpine Amador Calaveras Contra Costa El Dorado Fresno Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday

88 Alameda Alpine Amador Calaveras Contra Costa El Dorado Fresno Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Monday Monday Monday Monday Monday Monday Monday Monday Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday

89 Alameda Alpine Amador Calaveras Contra Costa El Dorado Fresno Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Friday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday

90 Glenn Inyo Kern Kings Los Angeles Madera Mariposa Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs

91 Glenn Inyo Kern Kings Los Angeles Madera Mariposa Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday

92 Glenn Inyo Kern Kings Los Angeles Madera Mariposa Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Mendocino Merced Mono Monterey Napa Plumas Riverside Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday

93 Mendocino Merced Mono Monterey Napa Plumas Riverside Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs

94 Mendocino Merced Mono Monterey Napa Plumas Riverside Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Tues/Wed/Thurs Tues/Wed/Thurs Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Holiday Holiday Holiday Holiday Holiday

95 Mendocino Merced Mono Monterey Napa Plumas Riverside Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Sacramento San Benito San Joaquin San Luis Obispo Santa Barbara Santa Clara Santa Cruz Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Monday Monday Monday Monday Monday Monday

96 Sacramento San Benito San Joaquin San Luis Obispo Santa Barbara Santa Clara Santa Cruz Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday

97 Sacramento San Benito San Joaquin San Luis Obispo Santa Barbara Santa Clara Santa Cruz Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday

98 Sacramento San Benito San Joaquin San Luis Obispo Santa Barbara Santa Clara Santa Cruz Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Holiday Holiday Holiday Holiday Solano Stanislaus Sutter Tulare Tuolumne Ventura Yolo Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Sunday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday Monday

99 Solano Stanislaus Sutter Tulare Tuolumne Ventura Yolo Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Monday Monday Monday Monday Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Tues/Wed/Thurs Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Friday Saturday Saturday Saturday

100 Solano Stanislaus Sutter Tulare Tuolumne Ventura Yolo Day of Week Hour LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH LD LM HH Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Saturday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday Holiday

101 Appendix C: Scaling procedures after DTIM processing C1. Block Diagram of Scaling Process: ldg (gas: heavy- and light-duty; diesel: light-duty) DTIM has 1 to 12 Source Classification Codes (SCC) that vary by species. For CO, NOx, SOx and PM species, DTIM only uses SCC=1 for the running exhaust emissions regardless of the fuel type and process. However, distribution of the running exhaust emissions according to the fuel type and process is needed. The following diagram explains how to distribute the running exhaust emissions for the light-duty gas. The running exhaust emissions are distributed to the catalyst cold exhaust, catalyst hot exhaust, non-catalyst cold exhaust, non-catalyst hot exhaust, catalyst bus and noncatalyst bus by using the corresponding emissions from EMFAC. Since there are no idle emissions in DTIM, surrogates are needed for the catalyst idle and non-catalyst idle. The surrogates for the catalyst idle and non-catalyst idle are catalyst hot exhaust, and non-catalyst hot exhaust, respectively. 100

102 C2. Block Diagram of Scaling Process: hdd (heavy-duty diesel) The following diagram explains how to distribute the running exhaust emissions for heavy-duty diesel. The running exhaust emissions are distributed to the diesel exhaust or diesel bus exhaust depending on the vehicle type by using the corresponding emissions from EMFAC. Since there are no idle emissions in DTIM, a surrogate is used. The surrogate for the diesel idle emissions is diesel exhaust or diesel bus exhaust, depending on the vehicle type. 101